diff --git a/BACKLOG.md b/BACKLOG.md
index 92027e5a..7c3d029a 100644
--- a/BACKLOG.md
+++ b/BACKLOG.md
@@ -31,7 +31,7 @@ CC-001/CC-002 were consumed by PR #24 fix bundle inline, with no standalone entr
 | CC-509 | ✅ closed 2026-07-22 | detached gate launch liveness：對 sandbox parent-death 早期死亡 fail-loud，提供 supervisor readiness／identity evidence | arch/gate | 2026-07-22 | pr:#440 | P2 | hygiene |
 | CC-510 | ✅ closed 2026-07-23 | Codex detached dispatch continuation：App Server callback、authenticated completion envelope 與 foreground fallback | arch/DX | 2026-07-23 | pr:#443 | P2 | design |
 | CC-511 | ⚠️ partial 2026-07-24 | ship publish authorization：Phase A current-tree authoritative full-suite 已交付；Phase B review-closure evidence 仍待 CC-515／CC-517 | release/gate | 2026-07-23 | pr:#446 | P1 | design |
-| CC-512 | 🔵 active | gate tier、execution mode、reviewer coverage 與 independence assurance 正交化 | ops/gate | 2026-07-23 | — | P1 | design |
+| CC-512 | ⚠️ partial 2026-07-27 | Slice A 已交付 coordinate sources／CLI resolution；machine envelope、evidence capture 與 shared verifier 仍待 Slice B／C | ops/gate | 2026-07-23 | — | P1 | design |
 | CC-513 | 🔵 active | canonical gate policy resolver：minimum tier、required reviewers、mode recommendation 與 downgrade audit | security/gate | 2026-07-23 | — | P1 | design |
 | CC-514 | 🔵 active | orthogonal delivery assurance map、machine-derived tables 與 feature/docs/high-risk recipes | docs/process | 2026-07-23 | — | P2 | design |
 | CC-515 | 🔵 active | gate artifact immutable subject、freshness 與 consumer applicability shared verifier | arch/gate | 2026-07-23 | — | P1 | design |
@@ -1616,47 +1616,114 @@ authorization。
 
 ---
 
-## CC-512 — tier／mode／coverage／independence assurance 正交化 🔵 active
-
-**Problem**: runtime 原本就將 tier detection、reviewer selection 與
-`SEQUENTIAL=true|false` 分開，但文件曾把 `full` 描述成 parallel cross-context +
-五個 reviewer，造成 tier、execution topology、實際 coverage 與 independence
-assurance 相互暗示。`full + sequential` 與 `express + parallel` 本應合法，artifact
-卻沒有足夠 metadata 防止錯誤宣稱。
+## CC-512 — tier／mode／pass／coverage／independence assurance 正交化 ⚠️ partial 2026-07-27
+
+**Problem**: runtime 雖已將 tier detection、reviewer selection 與
+`SEQUENTIAL=true|false` 分開，但目前仍有三個 truth gap：
+
+1. `--reviewers` 在沒有 explicit tier 時會把 `TIER` 改成 `targeted`，把 remediation
+   delta／review scope 誤分類成 rigor tier；`--targeted` 又只是 `--reviewers` alias，
+   無法引用 initial review。
+2. `express|standard|full` 實際主要選擇預設 reviewer 清單，文件卻把 `full` 寫成
+   parallel + 五 reviewer + 較高 assurance。requested intent、resolved defaults 與
+   actual evidence 沒有分開。
+3. final Markdown frontmatter 由 reviewer／synthesis session 產生；
+   `gate_result_verify` 只驗 `Final:` 唯一性與 frontmatter/body parity，不能證明實際
+   mode、selected reviewers、skipped reviewers 或 session independence。
+
+**Assurance coordinates**:
+
+1. **Tier** 是 rigor intent/preset，closed enum 僅
+   `express|standard|full`。記錄 `tier.requested: auto|<tier>` 與
+   `tier.resolved: <tier>`；tier table 可提供 default reviewers／evidence floor，
+   但不得把 default 當 actual coverage 或 policy authorization。
+2. **Mode** 只描述 execution topology。記錄
+   `mode.requested: default|sequential|parallel` 與
+   `mode.resolved: sequential|parallel`；sequential =
+   `combined-session`，parallel = `per-reviewer-sessions` + synthesis。
+3. **Pass kind** 獨立為 `initial|targeted`。Targeted 是 remediation delta review，
+   不是 tier；必須記錄 initial gate result reference 與 inherited／requested
+   coverage basis，不能冒充 initial comprehensive review。
+4. **Coverage** 分開記錄 requested、selected、skipped reviewer sets；tier defaults、
+   CLI override 與實際 dispatch 結果不得互相替代。
+5. **Independence** 分開記錄 implementation-context isolation、
+   reviewer topology、`per_reviewer_independent` 與 machine-captured session
+   evidence/status。缺 evidence 時只能標 `unverified|unavailable`，不能宣稱 verified。
 
 **Requirement**:
 
-1. 建立分離的 machine-readable policy tables：
-   - tier table：express/standard/full/targeted 的 default reviewers 與 evidence
-     floor；
-   - mode table：sequential combined session、parallel per-reviewer sessions +
-     synthesis 的 execution topology。
-   不得建立 `full => parallel` 的合併 table。
-2. requested/resolved 值分開記錄：
-   `tier.requested/resolved`、`mode.requested/resolved`。Tier 不改 mode，parallel
-   不提升 tier；express/standard/full/targeted × sequential/parallel 都是合法組合，
-   subject/policy constraints 另由 [[CC-513]] 判斷。
-3. artifact 明列實際 `coverage.reviewers`／`coverage.skipped`，以及 independence：
-   implementation-context isolation、`combined-session|per-reviewer-sessions`、
-   `per_reviewer_independent: true|false` 與可驗證 session evidence。不能由 tier、
-   mode 名稱或 reviewer 數量推論另一維度。
-4. targeted 結果必須標示 delta-specific 與引用的 initial coverage；不能呈現為完整
-   initial review。Sequential review 仍可證明與 implementation session 隔離，只是
-   reviewer 彼此共享 combined session。
-5. validator 拒絕／標 invalid：宣稱 parallel 卻無個別 session evidence、宣稱
-   per-reviewer independence 但實際 combined session、express+parallel 被呈現成
-   full coverage、full+sequential 被呈現成 parallel、targeted 被呈現成 initial full。
-6. CLI help、result frontmatter、review-model、commands、skills 與測試都由同一
-   machine source 派生或引用 bounded generated markers；README 只提供 pointer。
-
-**Done-when**: 任一 gate artifact 都能獨立回答「要求的深度、實際拓撲、實際 reviewer
-coverage、取得何種 independence」，所有合法組合 round-trip，錯誤 assurance claims
-被 validator 擋下。
-
-**Non-goals**: 不讓 full 自動 parallel；不讓 parallel 自動 full；不決定 risk-based
-minimum floor（見 [[CC-513]]）；不新增另一種 gate。
-
-**Cross-link**: [[CC-515]]、[[CC-518]]、[[CC-519]]、`docs/review-model.md`。
+1. 建立分離的 portable machine sources（預定
+   `core/policy/gate-tiers.tsv`、`core/policy/gate-modes.tsv`、
+   `core/policy/gate-pass-kinds.tsv`）；repo-layout runtime 直接讀 source，
+   standalone/copy-mode fallback 使用 bounded generated snapshot + freshness ratchet，
+   不手寫第二份 policy。三表分別擁有 defaults／topology／initial-reference requirement，
+   不建立合併 profile。
+2. CLI canonicalize：
+   - 新增 `--mode sequential|parallel`；既有 `--parallel`／`--sequential` 為
+     compatibility spelling，互相衝突時 fail closed。
+   - `--reviewers` 只覆蓋 requested coverage，不再改 tier。
+   - `--targeted <reviewers>` 表示 `pass.kind=targeted`，不再只是 alias；必須搭配
+     `--initial-result <path>`。Initial-result 的結構存在性在本票驗，subject freshness
+     與 applicability 留給 [[CC-515]]。
+   - omitted tier/mode/pass 分別 resolve 為 `auto`→detected tier、
+     `default`→sequential、`initial`。
+3. Gate shell 在 dispatch 前決定 coordinates，並在 dispatch／wait 時機械擷取實際
+   reviewer/synthesis session evidence；reviewer LLM 不得自行宣稱 tier、mode、
+   coverage 或 independence。
+4. 新 producer 寫 `pr_gate_result_v2` Markdown + `gate_assurance_v1` machine-owned
+   JSON sidecar；Markdown只保留human findings與bounded relative pointer。Envelope
+   至少含上述五組coordinates、actual reviewer dispatch outcome與evidence status；
+   artifact relocation、explicit `--output`、foreground/detached、
+   sequential/parallel及copy-mode都須byte-/meaning-parity。
+5. 擴充 shared verifier 只判斷本票擁有的
+   **structural + claim consistency**：enum、requested/resolved、selected/skipped
+   partition、mode/topology/session evidence、pass/initial reference及 Markdown pointer
+   parity。Stable repo subject、digest/freshness與 consumer policy applicability仍由
+   [[CC-515]]；risk-based floor仍由 [[CC-513]]。
+6. `pr_gate_result_v1` 維持 legacy structural verification，但明示
+   `assurance: unavailable`，不能被新 consumer 當完整 evidence；新 producer 不再
+   產生 v1。不得讓舊 artifact 因缺新欄位被誤報 forged。
+7. CLI help、result contract、`docs/review-model.md`、commands、skills與測試引用相同
+   machine source或 bounded generated markers；README只保留 pointer。修正既有
+   `full => parallel` rigor-tier敘述，但不把本票號寫入 operational docs。
+
+**Delivery slices（同一 ticket；不得提前宣稱完成）**:
+
+1. **A — coordinate sources + CLI resolution（✅ delivered 2026-07-27）**：
+   三份 canonical TSV、repo-layout direct load、copy-mode bounded snapshot +
+   parity ratchet、canonical `--mode`、targeted initial reference、closed
+   invalid/conflicting inputs，以及 dispatch brief 中的 requested/resolved／coverage
+   coordinates。`--tier full --reviewers critic` 保留 full intent + critic-only
+   selection；新 producer／verifier 尚未交付。
+2. **B — machine-owned envelope + evidence capture（pending）**：sequential combined session、
+   parallel per-reviewer/synthesis sessions、targeted initial reference、copy-mode
+   truthful degradation。
+3. **C — verifier + remaining parity ratchets（pending）**：claim consistency、v1 legacy
+   classification、result/help/docs parity、affected-test mapping。
+
+**Done-when**:
+
+- `express|standard|full × sequential|parallel × initial|targeted` 的合法矩陣可
+  round-trip；targeted fixtures皆帶 initial reference。
+- `full+sequential` 不宣稱 parallel、`express+parallel` 不宣稱 full coverage；
+  `--tier full --reviewers critic` 可誠實產生「full intent + critic-only actual
+  coverage」，是否 policy-sufficient 留給後續 verifier。
+- Validator 能抓到 missing/duplicate reviewer partition、targeted 無 initial、
+  parallel 無 per-session evidence、combined session 冒充 per-reviewer independent、
+  LLM frontmatter與machine envelope不一致、copy-mode冒充 verified，以及v1被誤當新
+  assurance。
+
+**Non-goals**: 不在本票決定 sensitive-path／risk-based minimum floor（[[CC-513]]）；
+不驗 artifact subject/freshness/publish applicability（[[CC-515]]）；不建立 scope
+manifest或finding schema（[[CC-518]]／[[CC-519]]）；不新增 gate kind、workflow
+engine、FSM或 mandatory parallel policy。
+
+**Dependencies**: 無 hard implementation dependency；本票先鎖定 coordinates，
+[[CC-513]]再產 policy resolution，[[CC-515]]再把 structural evidence 與
+subject/freshness/applicability接起來。
+
+**Cross-link**: [[CC-513]]、[[CC-515]]、[[CC-518]]、[[CC-519]]、
+`docs/review-model.md`。
 
 ---
 
@@ -1671,8 +1738,8 @@ generic 使用者會被不必要強制，maintainer 路徑則可能漏掉必要
 **Requirement**:
 
 1. 建立單一可測 resolver，輸入 diff classification、trusted brief metadata、
-   generated/untracked/renamed paths、requested tier/mode/reviewers、repo policy 與
-   accepted-risk override；輸出：
+   generated/untracked/renamed paths、requested tier/mode/pass/reviewers、repo policy
+   與 accepted-risk override；輸出：
    `minimum_tier`、`required_reviewers`、`recommended_mode`、`required_mode|null`、
    `downgrade_allowed`、matched signals 與 override provenance。
 2. **Generic `pmctl gate` risk-based floor**：
@@ -1787,7 +1854,8 @@ manifest 與 remediation closure 共同依賴，屬 P1 evidence foundation。
    publish authorization 與未來 consumer 呼叫；不得各自 grep `Final:` 或自行重做
    repo identity/freshness 邏輯。
 4. final artifact 連結 preflight evidence digest/subject、[[CC-512]] resolved tier/
-   mode/coverage、[[CC-513]] policy resolution，以及 [[CC-518]] scope manifest。
+   mode/pass/coverage/independence、[[CC-513]] policy resolution，以及 [[CC-518]]
+   scope manifest。
    finalize 前重新計算 working subject；HEAD/tree drift 標 stale，不產生可重用
    current-subject authorization。
 5. 覆蓋 result copy/replay、different physical worktree same git subject、different
diff --git a/DECISIONS.md b/DECISIONS.md
index 42993831..f08cfdc8 100644
--- a/DECISIONS.md
+++ b/DECISIONS.md
@@ -7,11 +7,55 @@ H2 標題格式：## YYYY-MM-DD: <短描述>
 與 BACKLOG closure 對應的 entry，內文首行寫：Closes: BACKLOG.md#<PREFIX>-NNN
 -->
 
+## 2026-07-27: targeted-review-is-a-pass-kind-not-a-tier
+
+Relates: CC-512, CC-513, CC-515, CC-517, CC-519
+
+**Context**: 2026-07-27 的 runtime planning audit確認：`pr-gate.sh` 在提供
+`--reviewers`、未提供 explicit `--tier` 時直接把 tier 設成 `targeted`；
+`--targeted` 只是同一 coverage override 的 alias。同時 final Markdown frontmatter
+由 reviewer／synthesis session撰寫，shared verifier只驗 `Final:` parity，無法證明
+實際 tier、mode、reviewer coverage或session independence。這使
+2026-07-23 decision裡「targeted屬於tier」與「tier／coverage／subject必須正交」
+互相矛盾。
+
+**Decision**:
+
+1. Tier closed enum收斂為 `express|standard|full`，表示 rigor intent/preset；
+   requested與resolved分開，default reviewer/evidence floor不是actual coverage。
+2. Review pass獨立為 `initial|targeted`。Targeted表示remediation delta，必須引用
+   initial result；它可和任一tier、sequential/parallel mode組合，但不得呈現為initial
+   comprehensive coverage。
+3. `--reviewers`只設定requested coverage；`--targeted`不再只是alias，並要求
+   `--initial-result`。Mode新增canonical `--mode`，既有`--parallel`／`--sequential`
+   保留為compatibility spelling。
+4. Tier、mode、pass、coverage與independence由gate shell產生machine-owned assurance
+   envelope；LLM Markdown只承載findings與human summary。CC-512 verifier只驗
+   structural/claim consistency；CC-513擁有risk/policy resolution，CC-515擁有
+   subject validity/freshness/applicability。
+
+**Alternatives considered**: (a) 保留`targeted`為第四種tier——否決，scope/delta不是
+rigor depth，且會繼續讓coverage override改寫tier。(b) 只在Markdown增加更多欄位——
+否決，欄位仍由LLM自述，無法作machine assurance。(c) 等CC-515一次解決——否決，
+CC-515需要先有穩定coordinates，否則subject verifier會同時承擔CLI語意與policy來源。
+
+**Constraints introduced**: 新producer不得再產生把targeted當tier的artifact；
+`pr_gate_result_v1`只能被分類為legacy structural evidence，不能被新consumer當成
+完整assurance。不得以tier名稱推論actual reviewer coverage、mode或publish
+authorization；不得新增gate kind、workflow engine或FSM。本決策只取代
+2026-07-23 entry中「targeted屬tier」的部分，其餘六維正交與publish authorization
+決策維持有效。
+
+---
+
 ## 2026-07-23: gate-delivery-assurance-dimensions-are-orthogonal
 
 Relates: CC-511, CC-512, CC-513, CC-514, CC-515, CC-517, CC-518, CC-519,
 CC-520, CC-521
 
+> 2026-07-27 clarification：本條目的 `targeted`-as-tier 部分已由
+> `targeted-review-is-a-pass-kind-not-a-tier` 取代；其餘六維正交決策不變。
+
 **Context**: 現有 runtime 已將 tier detection、reviewer selection 與
 `SEQUENTIAL=true|false` 分開處理，但 review 文件與規畫曾把 `full`、五個 reviewers、
 parallel independence 與 publish-ready GO 描述成同一件事。這造成兩類漂移：
diff --git a/MILESTONES.md b/MILESTONES.md
index d94010da..77897732 100644
--- a/MILESTONES.md
+++ b/MILESTONES.md
@@ -101,11 +101,15 @@
 
 ### Phase 7 — evidence + policy foundations
 
+> 實作順序：CC-512 先鎖定 machine-owned assurance coordinates；CC-513 才能在同一
+> vocabulary 上產 policy resolution；CC-515 最後把 structural evidence 與 immutable
+> subject／freshness／consumer applicability 接起來。三者是不同責任，不合併成 profile。
+
 | 票 | 摘要 | 狀態 |
 |----|------|------|
-| CC-515 | immutable subject；artifact validity、subject freshness、policy applicability 三軸 shared verifier | 🔵 |
-| CC-512 | tier／mode／reviewer coverage／independence 正交化；full+sequential、express+parallel 均合法 | 🔵 |
+| CC-512 | Slice A coordinate sources／CLI resolution 已交付；machine-owned assurance envelope、evidence capture、verifier 仍待 B／C；targeted 不再是 tier | ⚠️ |
 | CC-513 | canonical resolver：minimum tier、required reviewers、recommended/required mode、generic vs maintainer policy 與 downgrade audit | 🔵 |
+| CC-515 | immutable subject；artifact validity、subject freshness、policy applicability 三軸 shared verifier | 🔵 |
 
 ### Phase 8 — existing gate structured evidence
 
diff --git a/commands/pr-gate.md b/commands/pr-gate.md
index a13338bd..494b93ec 100644
--- a/commands/pr-gate.md
+++ b/commands/pr-gate.md
@@ -1,6 +1,6 @@
 ---
 description: Run the tiered pre-PR review pipeline on the current branch.
-argument-hint: "[express|standard|full] [--targeted r1,r2] [--scope context] [--parallel]"
+argument-hint: "[express|standard|full] [--targeted r1,r2 --initial-result path] [--scope context] [--mode sequential|parallel]"
 ---
 
 Run the PR gate via `pmctl gate run`. The `runtime/bin/pr-gate.sh` script is the
@@ -9,15 +9,15 @@ internal implementation; `pmctl gate run` is the preferred invocation surface.
 **Sequential mode (default):** all reviewers run in one combined session.
 Low main-thread token cost (~5k dispatch + read result).
 
-**Parallel mode (`--parallel`):** each reviewer runs in its own independent session
+**Parallel mode (`--mode parallel`; `--parallel` is compatible):** each reviewer runs in its own independent session
 followed by a PM synthesis session. Higher token cost — use for auth/payment/migration
 paths or when reviewer independence matters.
 
 | Situation | Args |
 |---|---|
 | Routine code / seed / docs changes | _(none)_ |
-| Re-gate after fixing specific findings | `--targeted qa-tester,risk-reviewer` |
-| Auth / payment / migration / sensitive paths | `--parallel` |
+| Re-gate after fixing specific findings | `--targeted qa-tester,risk-reviewer --initial-result <path>` |
+| Auth / payment / migration / sensitive paths | `--mode parallel` |
 | Force a specific tier | `express` / `standard` / `full` |
 
 ## Step 1 - Invoke pmctl directly
@@ -78,8 +78,9 @@ implementation model or alternate executor availability cannot be determined.
 RAW_ARGS="${ARGUMENTS:-}"
 TIER_OVERRIDE=""
 TARGETED_REVIEWERS=""
+INITIAL_RESULT=""
 SCOPE_TOKENS=()
-PARALLEL=false
+GATE_MODE=""
 GATE_EXECUTOR="<gate_executor>"
 GATE_MODEL=""
 
@@ -105,13 +106,37 @@ while [[ "$idx" -lt "${#TOKENS[@]}" ]]; do
     --targeted)
       idx=$((idx + 1))
       TARGETED_REVIEWERS="${TOKENS[$idx]:-}"
+      [[ -n "$TARGETED_REVIEWERS" ]] || { echo "error: --targeted requires a reviewer list" >&2; exit 2; }
+      ;;
+    --initial-result)
+      idx=$((idx + 1))
+      INITIAL_RESULT="${TOKENS[$idx]:-}"
+      [[ -n "$INITIAL_RESULT" ]] || { echo "error: --initial-result requires a path" >&2; exit 2; }
       ;;
     --scope)
       idx=$((idx + 1))
       [[ -n "${TOKENS[$idx]:-}" ]] && SCOPE_TOKENS+=("${TOKENS[$idx]}")
       ;;
-    --parallel)
-      PARALLEL=true
+    --mode)
+      idx=$((idx + 1))
+      requested_mode="${TOKENS[$idx]:-}"
+      case "$requested_mode" in
+        sequential|parallel) ;;
+        *) echo "error: --mode requires sequential or parallel" >&2; exit 2 ;;
+      esac
+      if [[ -n "$GATE_MODE" && "$GATE_MODE" != "$requested_mode" ]]; then
+        echo "error: conflicting gate mode options" >&2
+        exit 2
+      fi
+      GATE_MODE="$requested_mode"
+      ;;
+    --parallel|--sequential)
+      requested_mode="${tok#--}"
+      if [[ -n "$GATE_MODE" && "$GATE_MODE" != "$requested_mode" ]]; then
+        echo "error: conflicting gate mode options" >&2
+        exit 2
+      fi
+      GATE_MODE="$requested_mode"
       ;;
     --executor)
       idx=$((idx + 1))
@@ -129,6 +154,15 @@ while [[ "$idx" -lt "${#TOKENS[@]}" ]]; do
   idx=$((idx + 1))
 done
 
+if [[ -n "$TARGETED_REVIEWERS" && -z "$INITIAL_RESULT" ]]; then
+  echo "error: --targeted requires --initial-result <path>" >&2
+  exit 2
+fi
+if [[ -z "$TARGETED_REVIEWERS" && -n "$INITIAL_RESULT" ]]; then
+  echo "error: --initial-result is only valid with --targeted" >&2
+  exit 2
+fi
+
 # Validate after parsing so this also rejects a missing substitution of the
 # <gate_executor> default, not only an invalid explicit --executor value.
 case "$GATE_EXECUTOR" in
@@ -148,9 +182,9 @@ SCOPE="${SCOPE_TOKENS[*]:-}"
 GATE_ARGS=(--cd "<work_dir>" --executor "$GATE_EXECUTOR")
 [[ -n "$GATE_MODEL" ]] && GATE_ARGS+=(--model "$GATE_MODEL")
 [[ -n "$TIER_OVERRIDE" ]] && GATE_ARGS+=(--tier "$TIER_OVERRIDE")
-[[ -n "$TARGETED_REVIEWERS" ]] && GATE_ARGS+=(--reviewers "$TARGETED_REVIEWERS")
+[[ -n "$TARGETED_REVIEWERS" ]] && GATE_ARGS+=(--targeted "$TARGETED_REVIEWERS" --initial-result "$INITIAL_RESULT")
 [[ -n "$SCOPE" ]] && GATE_ARGS+=(--scope "$SCOPE")
-[[ "$PARALLEL" == "true" ]] && GATE_ARGS+=(--parallel)
+[[ -n "$GATE_MODE" ]] && GATE_ARGS+=(--mode "$GATE_MODE")
 
 # Launch detached: this call is inline (NOT run_in_background) and returns in
 # well under a second once the supervisor is forked -- stdout prints exactly
diff --git a/commands/ship.md b/commands/ship.md
index 25be64fc..129bcac0 100644
--- a/commands/ship.md
+++ b/commands/ship.md
@@ -173,10 +173,11 @@ once the call returns.
     assertion or fixture adjustment, or a small guard/error-handling fix.
 
   Then re-run `pmctl gate run --executor <gate_executor> --cd "<work_dir>"
-  --lifecycle foreground --reviewers <reviewer,...>` (same literal-path
-  substitution as Step 3's first call — never `"$PWD"`; the `/pr-gate`
-  `--targeted` flag maps to this same `--reviewers` option) for the reviewers
-  whose territory the fix touched. Repeat.
+  --lifecycle foreground --targeted <reviewer,...> --initial-result
+  "<initial_gate_result_path>"` (same literal-path substitution as Step 3's
+  first call — never `"$PWD"`). The initial-result path is the comprehensive
+  gate result that opened this remediation loop; keep it explicit on every
+  targeted pass. Repeat for the reviewers whose territory the fix touched.
 
 **Stop the loop only when**:
 1. Step 0's check would have caught this but didn't — implementation revealed
diff --git a/core/README.md b/core/README.md
index 1e4244f7..3426c7d4 100644
--- a/core/README.md
+++ b/core/README.md
@@ -5,7 +5,8 @@ This directory contains the canonical PM-runtime data contract. **`core/` is def
 - `schema/`        — JSON Schema files (`.schema.json`) for the 8 first-class
                      entities: Task, Run, Event, Review, Decision, Brief,
                      Handover, ContextPack.
-- `policy/`        — declarative tables (YAML) for enums and state machines.
+- `policy/`        — declarative YAML/TSV tables for enums, presets, and state
+                     machines.
 - `state/`         — definition of the on-disk state-store layout
                      (`~/.local/share/pm-dispatch/state/`). **Definitions, not the writer.**
 - `context-pack/`  — source-interface contract for ContextPack assembly.
@@ -25,7 +26,7 @@ This directory contains the canonical PM-runtime data contract. **`core/` is def
 ## Dependency graph (acyclic, downward)
 
 ```
-core/schema/         → core/policy/  (via inline enum + YAML editing source-of-truth)
+core/schema/         → core/policy/  (via inline enum + declarative editing source-of-truth)
 core/state/          → core/schema/  (layout.yaml names which schema each path conforms to)
 core/context-pack/   → core/schema/  (source.interface.md references context-pack.schema.json)
 ```
@@ -33,7 +34,7 @@ core/context-pack/   → core/schema/  (source.interface.md references context-p
 `core/policy/` is a leaf — depends on nothing.
 `core/state/` does NOT import `core/policy/`.
 
-The designated writer module in `runtime/lib/state-writer.sh` is the sole manager of writes to `~/.local/share/pm-dispatch/state/` paths. It validates every durable Run, Event, Task, and Decision write against recursive object requirements, constants, primitive types, enums, and `if`/`then` conditionals declared in `core/schema/`, using `jq` only. Runtime enum consumers read the corresponding `core/policy/` YAML, while parity tests keep the inline schema enum mirrors synchronized. Full draft-07 validation remains a development/test concern rather than a runtime dependency.
+The designated writer module in `runtime/lib/state-writer.sh` is the sole manager of writes to `~/.local/share/pm-dispatch/state/` paths. It validates every durable Run, Event, Task, and Decision write against recursive object requirements, constants, primitive types, enums, and `if`/`then` conditionals declared in `core/schema/`, using `jq` only. Runtime enum consumers read the corresponding declarative source under `core/policy/`, while parity tests keep inline schema mirrors and generated portability snapshots synchronized. Full draft-07 validation remains a development/test concern rather than a runtime dependency.
 
 ## Schema versioning
 
diff --git a/core/policy/gate-modes.tsv b/core/policy/gate-modes.tsv
new file mode 100644
index 00000000..96ffc411
--- /dev/null
+++ b/core/policy/gate-modes.tsv
@@ -0,0 +1,4 @@
+# Gate execution topology. Mode does not imply tier or reviewer coverage.
+mode	topology	synthesis	is_default
+sequential	combined-session	inline	true
+parallel	per-reviewer-sessions	separate-session	false
diff --git a/core/policy/gate-pass-kinds.tsv b/core/policy/gate-pass-kinds.tsv
new file mode 100644
index 00000000..7296bb61
--- /dev/null
+++ b/core/policy/gate-pass-kinds.tsv
@@ -0,0 +1,4 @@
+# Gate review-pass semantics. Pass kind does not imply tier or execution mode.
+pass_kind	scope	requires_initial_result	is_default
+initial	comprehensive	false	true
+targeted	remediation-delta	true	false
diff --git a/core/policy/gate-tiers.tsv b/core/policy/gate-tiers.tsv
new file mode 100644
index 00000000..eb1c8dba
--- /dev/null
+++ b/core/policy/gate-tiers.tsv
@@ -0,0 +1,5 @@
+# Gate rigor presets. Defaults do not assert actual coverage or execution topology.
+tier	default_reviewers	evidence_floor
+express	critic,qa-tester	reviewer-verdicts
+standard	critic,qa-tester,architecture-reviewer	reviewer-verdicts
+full	critic,qa-tester,architecture-reviewer,security-reviewer,risk-reviewer	reviewer-verdicts
diff --git a/core/schema/review.schema.json b/core/schema/review.schema.json
index 2776aa18..756ceb1a 100644
--- a/core/schema/review.schema.json
+++ b/core/schema/review.schema.json
@@ -1,6 +1,6 @@
 {
   "title": "Review",
-  "description": "A PR-gate review verdict, optionally with per-reviewer findings. `tier` mirrors /pr-gate tier names (express/standard/full/targeted). `findings_path` points at .gate-results/<gate-id>/<reviewer>.{json,md} — Review row is the INDEX, not the data.",
+  "description": "A PR-gate review verdict, optionally with per-reviewer findings. Current rigor tiers come from core/policy/gate-tiers.tsv. The schema_version 1 `targeted` tier value is retained only to validate legacy payloads; targeted remediation is now a pass kind. `findings_path` points at .gate-results/<gate-id>/<reviewer>.{json,md} — Review row is the INDEX, not the data.",
   "type": "object",
   "required": ["id", "schema_version", "task_id", "tier", "verdict", "findings", "created_ts"],
   "properties": {
@@ -20,7 +20,8 @@
       "description": "Optional link to the Run that triggered the gate."
     },
     "tier": {
-      "enum": ["express", "standard", "full", "targeted"]
+      "enum": ["express", "standard", "full", "targeted"],
+      "description": "Current editing source: core/policy/gate-tiers.tsv. `targeted` is a schema_version 1 compatibility value and must not be emitted by new producers."
     },
     "verdict": {
       "enum": ["GO", "NO-GO", "pending"]
diff --git a/docs/review-model.md b/docs/review-model.md
index cd903a81..8d4f883f 100644
--- a/docs/review-model.md
+++ b/docs/review-model.md
@@ -143,19 +143,33 @@ When an artifact is missing — no `conceptual_map`, no `self_verify cmd:`, no i
 
 ---
 
-## pr-gate rigor tiers
+## pr-gate assurance coordinates
 
 The `/pr-gate` `--tier` flag selects the **rigor level** required for this change — not just the number of reviewers. Choose based on `architecture_impact` and blast radius:
 
-| Tier | When | Rigor |
+| Tier | When | Default reviewer coverage |
 |---|---|---|
-| `express` | hotfix, docs-only, `architecture_impact: none` | machine verification + combined session (critic + qa) |
+| `express` | hotfix, docs-only, `architecture_impact: none` | critic + qa |
 | `standard` | feature, `architecture_impact: minor` | conceptual map required + critic + qa + architecture-reviewer |
-| `full` | architectural change, `architecture_impact: major`, sensitive path | parallel cross-context sessions + security + risk hard gates + synthesis |
+| `full` | architectural change, `architecture_impact: major`, sensitive path | critic + qa + architecture-reviewer + security + risk |
 
 **Tier suggestion**: when a `--brief` is passed to `pr-gate.sh`, it reads `architecture_impact` from the brief and emits an advisory to stderr before dispatch if the auto-detected tier is lower than the impact level implies. The user-selected or auto-detected tier always takes precedence; the advisory is informational only and does not block or alter the tier.
 
-**Tier vs. reviewer count**: `express` is not "fewer reviewers" — it is "this change has bounded impact and does not need architecture-level judgment." `full` is not "more reviewers" — it is "this change has wide blast radius and needs independent parallel review with hard security/risk gates."
+**Tier, mode, pass kind, and coverage are independent**:
+
+- Tier records rigor intent and supplies default reviewer coverage. `--reviewers`
+  may override the selected coverage without rewriting the tier.
+- Mode records execution topology. The default is `sequential`; select
+  `--mode parallel` when separate reviewer sessions and synthesis are required.
+  A `full` tier does not select parallel mode by itself.
+- Pass kind records whether the review is initial or a remediation-delta
+  targeted pass. `--targeted <reviewers>` requires
+  `--initial-result <path>` and is not a tier alias.
+
+The portable policy sources are
+[`core/policy/gate-tiers.tsv`](../core/policy/gate-tiers.tsv),
+[`core/policy/gate-modes.tsv`](../core/policy/gate-modes.tsv), and
+[`core/policy/gate-pass-kinds.tsv`](../core/policy/gate-pass-kinds.tsv).
 
 ---
 
diff --git a/runtime/bin/pr-gate.sh b/runtime/bin/pr-gate.sh
index 133e5895..4cabee9c 100755
--- a/runtime/bin/pr-gate.sh
+++ b/runtime/bin/pr-gate.sh
@@ -17,6 +17,168 @@ trap '' PIPE
 # shellcheck disable=SC2059  # printf passthrough wrapper: the caller owns the format string
 say() { printf "$@" 2>/dev/null || true; }
 
+# Portable policy reader for the gate assurance coordinates.
+#
+# Repo-layout runs read the canonical TSV files under core/policy/. A copied
+# gate may not carry that tree, so it falls back to the bounded generated
+# snapshot below. tests/shell/test-pr-gate.sh compares each heredoc byte-for-
+# byte with its canonical source so this fallback cannot drift silently.
+_gate_assurance_policy_snapshot() {
+  case "${1:-}" in
+    tiers)
+      # BEGIN GENERATED from core/policy/gate-tiers.tsv
+      cat <<'GATE_ASSURANCE_TIERS_TSV'
+# Gate rigor presets. Defaults do not assert actual coverage or execution topology.
+tier	default_reviewers	evidence_floor
+express	critic,qa-tester	reviewer-verdicts
+standard	critic,qa-tester,architecture-reviewer	reviewer-verdicts
+full	critic,qa-tester,architecture-reviewer,security-reviewer,risk-reviewer	reviewer-verdicts
+GATE_ASSURANCE_TIERS_TSV
+      # END GENERATED from core/policy/gate-tiers.tsv
+      ;;
+    modes)
+      # BEGIN GENERATED from core/policy/gate-modes.tsv
+      cat <<'GATE_ASSURANCE_MODES_TSV'
+# Gate execution topology. Mode does not imply tier or reviewer coverage.
+mode	topology	synthesis	is_default
+sequential	combined-session	inline	true
+parallel	per-reviewer-sessions	separate-session	false
+GATE_ASSURANCE_MODES_TSV
+      # END GENERATED from core/policy/gate-modes.tsv
+      ;;
+    pass-kinds)
+      # BEGIN GENERATED from core/policy/gate-pass-kinds.tsv
+      cat <<'GATE_ASSURANCE_PASS_KINDS_TSV'
+# Gate review-pass semantics. Pass kind does not imply tier or execution mode.
+pass_kind	scope	requires_initial_result	is_default
+initial	comprehensive	false	true
+targeted	remediation-delta	true	false
+GATE_ASSURANCE_PASS_KINDS_TSV
+      # END GENERATED from core/policy/gate-pass-kinds.tsv
+      ;;
+    *)
+      printf 'pr-gate: unknown assurance policy table: %s\n' "${1:-empty}" >&2
+      return 2
+      ;;
+  esac
+}
+
+_gate_assurance_policy_filename() {
+  case "${1:-}" in
+    tiers) printf 'gate-tiers.tsv\n' ;;
+    modes) printf 'gate-modes.tsv\n' ;;
+    pass-kinds) printf 'gate-pass-kinds.tsv\n' ;;
+    *) return 2 ;;
+  esac
+}
+
+_gate_assurance_policy_path() {
+  local filename candidate
+  filename="$(_gate_assurance_policy_filename "${1:-}")" || return 2
+  for candidate in \
+    "$SCRIPT_DIR/../../core/policy/$filename" \
+    "$SCRIPT_DIR/core/policy/$filename"
+  do
+    if [[ -r "$candidate" ]]; then
+      printf '%s\n' "$candidate"
+      return 0
+    fi
+  done
+  return 1
+}
+
+_gate_assurance_policy_emit() {
+  local path
+  if path="$(_gate_assurance_policy_path "${1:-}")"; then
+    cat "$path"
+  else
+    _gate_assurance_policy_snapshot "${1:-}"
+  fi
+}
+
+# _gate_assurance_policy_lookup <table> <key-column> <key> <value-column>
+# Requires exactly one matching row and rejects malformed/duplicate tables.
+_gate_assurance_policy_lookup() {
+  local table="${1:-}" key_column="${2:-}" key="${3:-}" value_column="${4:-}"
+  [[ $# -eq 4 ]] || return 2
+  _gate_assurance_policy_emit "$table" | awk -F '\t' \
+    -v key_name="$key_column" -v wanted="$key" -v value_name="$value_column" '
+      /^[[:space:]]*#/ || /^[[:space:]]*$/ { next }
+      !header_seen {
+        header_seen=1
+        header_width=NF
+        for (i=1; i<=NF; i++) {
+          sub(/\r$/, "", $i)
+          if ($i == key_name) key_index=i
+          if ($i == value_name) value_index=i
+        }
+        if (!key_index || !value_index) { malformed=1; exit }
+        next
+      }
+      {
+        sub(/\r$/, "", $NF)
+        if (NF != header_width) { malformed=1; exit }
+        if ($(key_index) == wanted) {
+          matches++
+          result=$(value_index)
+        }
+      }
+      END {
+        if (malformed || !header_seen || matches != 1 || result == "") exit 2
+        print result
+      }
+    '
+}
+
+# _gate_assurance_policy_values <table> <key-column>
+# Emits a unique, non-empty closed enum in source order.
+_gate_assurance_policy_values() {
+  local table="${1:-}" key_column="${2:-}"
+  [[ $# -eq 2 ]] || return 2
+  _gate_assurance_policy_emit "$table" | awk -F '\t' -v key_name="$key_column" '
+    /^[[:space:]]*#/ || /^[[:space:]]*$/ { next }
+    !header_seen {
+      header_seen=1
+      header_width=NF
+      for (i=1; i<=NF; i++) {
+        sub(/\r$/, "", $i)
+        if ($i == key_name) key_index=i
+      }
+      if (!key_index) { malformed=1; exit }
+      next
+    }
+    {
+      sub(/\r$/, "", $NF)
+      value=$(key_index)
+      if (NF != header_width || value == "" || seen[value]++) {
+        malformed=1
+        exit
+      }
+      values[++count]=value
+    }
+    END {
+      if (malformed || !header_seen || count == 0) exit 2
+      for (i=1; i<=count; i++) print values[i]
+    }
+  '
+}
+
+_gate_set_mode_requested() {
+  local candidate="$1" spelling="$2"
+  if [[ "$MODE_OPTION_SEEN" == false ]]; then
+    MODE_REQUESTED="$candidate"
+    MODE_OPTION_SEEN=true
+    MODE_OPTION_SPELLING="$spelling"
+    return 0
+  fi
+  if [[ "$MODE_REQUESTED" != "$candidate" ]]; then
+    printf 'Error: conflicting gate mode options: %s requested %s, but %s requested %s\n' \
+      "$MODE_OPTION_SPELLING" "$MODE_REQUESTED" "$spelling" "$candidate" >&2
+    return 2
+  fi
+  return 0
+}
+
 # verify_reviewer_artifact_hashes <hash_cmd> <name> <path> <baseline> [...]
 # Print every reviewer whose artifact differs from its captured baseline.
 verify_reviewer_artifact_hashes() {
@@ -45,6 +207,7 @@ _kill_process_tree() {
   kill -"$_sig" "$_pid" 2>/dev/null || true
 }
 
+# pr-gate-help:start
 # pr-gate.sh -- PR-gate review via a dispatched session
 #
 # DEFAULT (single-session / sequential):
@@ -52,7 +215,7 @@ _kill_process_tree() {
 #   Lower token cost. All reviewer findings appear in a single output file.
 #   Use this for most routine changes.
 #
-# MULTI-SESSION (--parallel):
+# MULTI-SESSION (--mode parallel; --parallel is compatible):
 #   Each reviewer runs in its own INDEPENDENT dispatch session, followed by a
 #   separate PM synthesis session. Reviewers share no context window, which
 #   eliminates anchoring bias across reviewers.
@@ -69,9 +232,11 @@ _kill_process_tree() {
 # Options:
 #   --cd <dir>           working directory (required)
 #   --tier <tier>        express|standard|full -- overrides auto-detection
+#   --mode <mode>        sequential|parallel execution topology (default: sequential)
 #   --brief <file>       dispatch brief for this change; architecture_impact field informs tier suggestion
-#   --reviewers <list>   comma-separated names -- overrides tier default (targeted re-gate)
-#   --targeted <list>    alias for --reviewers (matches /pr-gate skill vocabulary)
+#   --reviewers <list>   comma-separated requested coverage; does not change tier or pass kind
+#   --targeted <list>    remediation-delta pass over these reviewers; requires --initial-result
+#   --initial-result <f> initial gate result referenced by a --targeted pass; relative to --cd
 #   --reviewer-dir <dir> explicit reviewer-definition source; defaults to the repo-owned agents/ directory
 #   --scope <text>       context hint passed into the review brief
 #   --base <branch>      base branch for diff (default: origin/HEAD → main)
@@ -94,8 +259,8 @@ _kill_process_tree() {
 #                        this to dial reasoning depth up/down without switching models.
 #   --isolation <level>  isolation level: none|read-only|workspace-write|workspace-network|sandboxed
 #   --timeout <secs>     dispatch timeout per session (default: 1200)
-#   --parallel           multi-session: one dispatch per reviewer + synthesis (higher token cost)
-#   --sequential         alias for default single-session mode (kept for backward compatibility)
+#   --parallel           compatibility spelling for --mode parallel
+#   --sequential         compatibility spelling for --mode sequential
 #   --allow-hooks        execute repo-local .pm-dispatch hook scripts (trusted branches only)
 #   --allow-dirty        review the working tree as-is instead of failing on a dirty tree atop committed changes
 #   --override-file <f>  inject accepted-risk overrides into every reviewer brief; auto-discovered
@@ -120,18 +285,26 @@ _kill_process_tree() {
 #   --test-timeout <s>   timeout for --test-cmd (default: 1800), decoupled from --timeout so a slow
 #                        test suite can never cause a reviewer dispatch session to time out.
 #   --skip-preflight-tests   force-disable the pre-flight test check even if --test-cmd is passed.
+# pr-gate-help:end
 
 WORK_DIR=""
 GATE_RUN_DIR_OVERRIDE=""   # out-of-repo artifact root; set via --run-dir from pmctl-gate
 TIER_OVERRIDE=""
+TIER_REQUESTED="auto"
 REVIEWERS_OVERRIDE=""
+REVIEWERS_OPTION_SOURCE=""
+MODE_REQUESTED="default"
+MODE_OPTION_SEEN=false
+MODE_OPTION_SPELLING=""
+PASS_KIND_REQUESTED="initial"
+INITIAL_RESULT_INPUT=""
+INITIAL_RESULT_OPTION_SEEN=false
 REVIEWER_DIR_OVERRIDE=""
 SCOPE=""
 BASE_OVERRIDE=""
 HEAD_OVERRIDE=""
 OUTPUT_OVERRIDE=""
 TIMEOUT="1200"
-SEQUENTIAL=true   # default: sequential (lower token cost)
 EXECUTOR_OPTION="auto"
 ALLOW_HOOKS=false   # hooks require explicit --allow-hooks opt-in (security)
 ALLOW_DIRTY=false   # gate refuses a dirty tree atop committed changes unless this opt-in
@@ -156,12 +329,42 @@ SKIP_PREFLIGHT_TESTS=false
 
 while [[ $# -gt 0 ]]; do
   case "$1" in
-    --cd)         WORK_DIR="$2";           shift 2;;
-    --run-dir)    GATE_RUN_DIR_OVERRIDE="$2"; shift 2;;
-    --tier)       TIER_OVERRIDE="$2";      shift 2;;
+    --cd)
+      [[ $# -ge 2 && -n "$2" && "$2" != --* ]] || { printf 'Error: --cd requires a directory\n' >&2; exit 2; }
+      WORK_DIR="$2"; shift 2;;
+    --run-dir)
+      [[ $# -ge 2 && -n "$2" && "$2" != --* ]] || { printf 'Error: --run-dir requires a directory\n' >&2; exit 2; }
+      GATE_RUN_DIR_OVERRIDE="$2"; shift 2;;
+    --tier)
+      [[ $# -ge 2 && -n "$2" && "$2" != --* ]] || { printf 'Error: --tier requires a value\n' >&2; exit 2; }
+      TIER_OVERRIDE="$2"; TIER_REQUESTED="$2"; shift 2;;
+    --mode)
+      [[ $# -ge 2 && -n "$2" && "$2" != --* ]] || { printf 'Error: --mode requires sequential or parallel\n' >&2; exit 2; }
+      _gate_set_mode_requested "$2" "--mode" || exit 2
+      shift 2;;
     --brief)      BRIEF_FILE="$2";         shift 2;;
-    --reviewers)  REVIEWERS_OVERRIDE="$2"; shift 2;;
-    --targeted)   REVIEWERS_OVERRIDE="$2"; shift 2;;   # alias: /pr-gate skill + script comments say "targeted"
+    --reviewers)
+      [[ $# -ge 2 && -n "$2" && "$2" != --* ]] || { printf 'Error: --reviewers requires a reviewer list\n' >&2; exit 2; }
+      [[ -z "$REVIEWERS_OPTION_SOURCE" ]] || {
+        printf 'Error: --reviewers and --targeted may not be combined or repeated\n' >&2
+        exit 2
+      }
+      REVIEWERS_OVERRIDE="$2"; REVIEWERS_OPTION_SOURCE="--reviewers"; shift 2;;
+    --targeted)
+      [[ $# -ge 2 && -n "$2" && "$2" != --* ]] || { printf 'Error: --targeted requires a reviewer list\n' >&2; exit 2; }
+      [[ -z "$REVIEWERS_OPTION_SOURCE" ]] || {
+        printf 'Error: --reviewers and --targeted may not be combined or repeated\n' >&2
+        exit 2
+      }
+      REVIEWERS_OVERRIDE="$2"; REVIEWERS_OPTION_SOURCE="--targeted"
+      PASS_KIND_REQUESTED="targeted"; shift 2;;
+    --initial-result)
+      [[ $# -ge 2 && -n "$2" && "$2" != --* ]] || { printf 'Error: --initial-result requires a result path\n' >&2; exit 2; }
+      [[ "$INITIAL_RESULT_OPTION_SEEN" == false ]] || {
+        printf 'Error: --initial-result may only be provided once\n' >&2
+        exit 2
+      }
+      INITIAL_RESULT_INPUT="$2"; INITIAL_RESULT_OPTION_SEEN=true; shift 2;;
     --reviewer-dir)
       [[ $# -ge 2 ]] || { printf 'Error: --reviewer-dir requires a directory\n' >&2; exit 2; }
       REVIEWER_DIR_OVERRIDE="$2"; shift 2;;
@@ -184,8 +387,12 @@ while [[ $# -gt 0 ]]; do
       DISPATCH_EFFORT="$2"; shift 2;;
     --isolation)  DISPATCH_ISOLATION="$2"; shift 2;;
     --timeout)    TIMEOUT="$2";            shift 2;;
-    --parallel)   SEQUENTIAL=false;        shift;;
-    --sequential) SEQUENTIAL=true;         shift;;   # backward compat
+    --parallel)
+      _gate_set_mode_requested "parallel" "--parallel" || exit 2
+      shift;;
+    --sequential)
+      _gate_set_mode_requested "sequential" "--sequential" || exit 2
+      shift;;
     --allow-hooks) ALLOW_HOOKS=true;       shift;;
     --allow-dirty) ALLOW_DIRTY=true;       shift;;
     --override-file)
@@ -202,11 +409,19 @@ while [[ $# -gt 0 ]]; do
       TEST_TIMEOUT="$2";   shift 2;;
     --skip-preflight-tests) SKIP_PREFLIGHT_TESTS=true; shift;;
     -h|--help)
-      sed -n '2,99p' "$0" | sed 's/^# \{0,1\}//'
+      awk '
+        /^# pr-gate-help:start$/ { help=1; next }
+        /^# pr-gate-help:end$/ { exit }
+        help {
+          line=$0
+          sub(/^# ?/, "", line)
+          print line
+        }
+      ' "$0"
       exit 0;;
     *)
       printf 'Unknown arg: %s\n' "$1" >&2
-      printf 'Accepted: --cd --run-dir --tier --brief --reviewers|--targeted --reviewer-dir --scope --base --head --output --executor --model --effort --isolation --timeout --parallel --sequential --allow-hooks --allow-dirty --override-file --test-cmd --test-timeout --skip-preflight-tests (-h for help)\n' >&2
+      printf 'Accepted: --cd --run-dir --tier --mode --brief --reviewers --targeted --initial-result --reviewer-dir --scope --base --head --output --executor --model --effort --isolation --timeout --parallel --sequential --allow-hooks --allow-dirty --override-file --test-cmd --test-timeout --skip-preflight-tests (-h for help)\n' >&2
       exit 2;;
   esac
 done
@@ -284,6 +499,186 @@ else
     return 0
   }
 fi
+
+# ── Resolve assurance policy coordinates ─────────────────────────────────────
+# These are resolved before any executor routing or dispatch. The LLM receives
+# the resolved values as read-only context; it does not choose or infer them.
+if ! GATE_TIER_VALUES="$(_gate_assurance_policy_values tiers tier)"; then
+  printf 'Error: invalid gate tier policy source\n' >&2
+  exit 2
+fi
+if ! GATE_MODE_VALUES="$(_gate_assurance_policy_values modes mode)"; then
+  printf 'Error: invalid gate mode policy source\n' >&2
+  exit 2
+fi
+if ! GATE_PASS_KIND_VALUES="$(_gate_assurance_policy_values pass-kinds pass_kind)"; then
+  printf 'Error: invalid gate pass-kind policy source\n' >&2
+  exit 2
+fi
+
+if ! GATE_MODE_DEFAULT="$(_gate_assurance_policy_lookup modes is_default true mode)"; then
+  printf 'Error: gate mode policy must declare exactly one default\n' >&2
+  exit 2
+fi
+if ! GATE_PASS_KIND_DEFAULT="$(_gate_assurance_policy_lookup pass-kinds is_default true pass_kind)"; then
+  printf 'Error: gate pass-kind policy must declare exactly one default\n' >&2
+  exit 2
+fi
+if [[ "$GATE_MODE_DEFAULT" != "sequential" ]]; then
+  printf 'Error: gate mode policy default must remain sequential\n' >&2
+  exit 2
+fi
+if [[ "$GATE_PASS_KIND_DEFAULT" != "initial" ]]; then
+  printf 'Error: gate pass-kind policy default must remain initial\n' >&2
+  exit 2
+fi
+
+if [[ -n "$TIER_OVERRIDE" ]] \
+    && ! _gate_assurance_policy_lookup tiers tier "$TIER_OVERRIDE" default_reviewers >/dev/null; then
+  printf 'Error: --tier must be one of: %s (got: %s)\n' \
+    "$(printf '%s\n' "$GATE_TIER_VALUES" | awk 'BEGIN{ORS=" "} {print} END{print "\n"}' | sed 's/[[:space:]]*$//')" \
+    "$TIER_OVERRIDE" >&2
+  exit 2
+fi
+
+if [[ "$MODE_OPTION_SEEN" == false ]]; then
+  MODE_RESOLVED="$GATE_MODE_DEFAULT"
+else
+  MODE_RESOLVED="$MODE_REQUESTED"
+fi
+if ! MODE_TOPOLOGY="$(_gate_assurance_policy_lookup modes mode "$MODE_RESOLVED" topology)" \
+    || ! MODE_SYNTHESIS="$(_gate_assurance_policy_lookup modes mode "$MODE_RESOLVED" synthesis)"; then
+  printf 'Error: --mode must be one of: %s (got: %s)\n' \
+    "$(printf '%s\n' "$GATE_MODE_VALUES" | awk 'BEGIN{ORS=" "} {print} END{print "\n"}' | sed 's/[[:space:]]*$//')" \
+    "$MODE_RESOLVED" >&2
+  exit 2
+fi
+case "$MODE_TOPOLOGY:$MODE_SYNTHESIS" in
+  combined-session:inline) SEQUENTIAL=true ;;
+  per-reviewer-sessions:separate-session) SEQUENTIAL=false ;;
+  *)
+    printf 'Error: unsupported gate mode topology for %s: %s + %s\n' \
+      "$MODE_RESOLVED" "$MODE_TOPOLOGY" "$MODE_SYNTHESIS" >&2
+    exit 2
+    ;;
+esac
+
+PASS_KIND_RESOLVED="$PASS_KIND_REQUESTED"
+if ! PASS_SCOPE="$(_gate_assurance_policy_lookup pass-kinds pass_kind "$PASS_KIND_RESOLVED" scope)" \
+    || ! PASS_REQUIRES_INITIAL="$(_gate_assurance_policy_lookup pass-kinds pass_kind "$PASS_KIND_RESOLVED" requires_initial_result)"; then
+  printf 'Error: gate pass kind must be one of: %s (got: %s)\n' \
+    "$(printf '%s\n' "$GATE_PASS_KIND_VALUES" | awk 'BEGIN{ORS=" "} {print} END{print "\n"}' | sed 's/[[:space:]]*$//')" \
+    "$PASS_KIND_RESOLVED" >&2
+  exit 2
+fi
+case "$PASS_REQUIRES_INITIAL" in
+  true)
+    if [[ -z "$INITIAL_RESULT_INPUT" ]]; then
+      printf 'Error: --targeted requires --initial-result <path>\n' >&2
+      exit 2
+    fi
+    ;;
+  false)
+    if [[ -n "$INITIAL_RESULT_INPUT" ]]; then
+      printf 'Error: --initial-result is only valid with --targeted\n' >&2
+      exit 2
+    fi
+    ;;
+  *)
+    printf 'Error: invalid requires_initial_result value for pass kind %s: %s\n' \
+      "$PASS_KIND_RESOLVED" "$PASS_REQUIRES_INITIAL" >&2
+    exit 2
+    ;;
+esac
+
+INITIAL_RESULT_RESOLVED=""
+if [[ -n "$INITIAL_RESULT_INPUT" ]]; then
+  _initial_result_candidate="$INITIAL_RESULT_INPUT"
+  [[ "$_initial_result_candidate" == /* ]] \
+    || _initial_result_candidate="$WORK_DIR/$_initial_result_candidate"
+  if [[ ! -f "$_initial_result_candidate" || ! -r "$_initial_result_candidate" \
+      || ! -s "$_initial_result_candidate" ]]; then
+    printf 'Error: --initial-result must name a readable, non-empty file: %s\n' \
+      "$INITIAL_RESULT_INPUT" >&2
+    exit 2
+  fi
+  _initial_result_parent="$(cd "$(dirname "$_initial_result_candidate")" && pwd -P)" || {
+    printf 'Error: cannot resolve --initial-result parent: %s\n' "$INITIAL_RESULT_INPUT" >&2
+    exit 2
+  }
+  INITIAL_RESULT_RESOLVED="$_initial_result_parent/$(basename "$_initial_result_candidate")"
+  if ! gate_result_verify "$INITIAL_RESULT_RESOLVED" "" "targeted initial result"; then
+    printf 'Error: --initial-result is not a structurally valid gate result: %s\n' \
+      "$INITIAL_RESULT_INPUT" >&2
+    exit 2
+  fi
+  unset _initial_result_candidate _initial_result_parent
+fi
+
+# Build the closed reviewer vocabulary from tier defaults in source order.
+# This is a vocabulary only: a tier default is not actual selected coverage.
+ALL_REVIEWERS=""
+while IFS= read -r _policy_tier; do
+  _policy_reviewers="$(_gate_assurance_policy_lookup tiers tier "$_policy_tier" default_reviewers)" || {
+    printf 'Error: invalid default_reviewers for gate tier: %s\n' "$_policy_tier" >&2
+    exit 2
+  }
+  _gate_assurance_policy_lookup tiers tier "$_policy_tier" evidence_floor >/dev/null || {
+    printf 'Error: missing evidence_floor for gate tier: %s\n' "$_policy_tier" >&2
+    exit 2
+  }
+  for _policy_reviewer in $(printf '%s' "$_policy_reviewers" | tr ',' ' '); do
+    if [[ ! "$_policy_reviewer" =~ ^[a-z0-9][a-z0-9-]*$ ]]; then
+      printf 'Error: invalid reviewer name in gate tier policy: %s\n' "$_policy_reviewer" >&2
+      exit 2
+    fi
+    if [[ " $ALL_REVIEWERS " != *" $_policy_reviewer "* ]]; then
+      ALL_REVIEWERS="${ALL_REVIEWERS:+$ALL_REVIEWERS }$_policy_reviewer"
+    fi
+  done
+done <<< "$GATE_TIER_VALUES"
+unset _policy_tier _policy_reviewers _policy_reviewer
+if [[ -z "$ALL_REVIEWERS" ]]; then
+  printf 'Error: gate tier policy did not declare any reviewers\n' >&2
+  exit 2
+fi
+
+_gate_normalize_reviewer_list() {
+  local raw="$1" source_label="$2" normalized="" reviewer
+  if [[ -z "${raw//[[:space:]]/}" || "$raw" == ,* || "$raw" == *, || "$raw" == *,,* ]]; then
+    printf 'Error: %s requires a non-empty comma-separated reviewer list\n' "$source_label" >&2
+    return 2
+  fi
+  raw="${raw//,/ }"
+  for reviewer in $raw; do
+    if [[ " $ALL_REVIEWERS " != *" $reviewer "* ]]; then
+      printf 'Error: %s contains unknown reviewer %s (allowed: %s)\n' \
+        "$source_label" "$reviewer" "$ALL_REVIEWERS" >&2
+      return 2
+    fi
+    if [[ " $normalized " == *" $reviewer "* ]]; then
+      printf 'Error: %s contains duplicate reviewer: %s\n' "$source_label" "$reviewer" >&2
+      return 2
+    fi
+    normalized="${normalized:+$normalized }$reviewer"
+  done
+  [[ -n "$normalized" ]] || return 2
+  printf '%s\n' "$normalized"
+}
+
+_gate_policy_source_count=0
+for _gate_policy_table in tiers modes pass-kinds; do
+  if _gate_assurance_policy_path "$_gate_policy_table" >/dev/null; then
+    _gate_policy_source_count=$((_gate_policy_source_count + 1))
+  fi
+done
+case "$_gate_policy_source_count" in
+  3) GATE_ASSURANCE_POLICY_SOURCE="canonical" ;;
+  0) GATE_ASSURANCE_POLICY_SOURCE="generated-snapshot" ;;
+  *) GATE_ASSURANCE_POLICY_SOURCE="mixed" ;;
+esac
+unset _gate_policy_source_count _gate_policy_table
+
 EXECUTOR_ROUTER_PATH="$SCRIPT_DIR/../lib/executor-router.sh"
 if [[ -r "$EXECUTOR_ROUTER_PATH" ]]; then
   # shellcheck source=runtime/lib/executor-router.sh
@@ -851,24 +1246,31 @@ fi
 
 # ── Detect tier ───────────────────────────────────────────────────────────────
 if [[ -n "$TIER_OVERRIDE" ]]; then
-  TIER="$TIER_OVERRIDE"
-elif [[ -n "$REVIEWERS_OVERRIDE" ]]; then
-  TIER="targeted"
+  TIER_RESOLVED="$TIER_OVERRIDE"
 else
   NON_DOCS=$(printf '%s\n' "$DIFF_FILES" | grep -vE '\.(md|jsonl|txt)$|^\.gitignore$|^audits/|^docs/' || true)
   SENSITIVE_HIT=$(printf '%s\n' "$DIFF_FILES" | { grep -iE '(^|[/_.-])(auth|oauth|jwt|session|secret|password|token|credential|cors|csrf|webhook|sudo|ssh|payment|billing)([/_.-]|$)|(^|/)migrations?/|^\.github/' || true; } | wc -l)
 
   if [[ -z "$NON_DOCS" ]]; then
-    TIER=express
+    TIER_RESOLVED=express
   elif [[ "$SENSITIVE_HIT" -gt 0 || "$LINES" -gt 500 ]]; then
-    TIER=full
+    TIER_RESOLVED=full
   elif [[ "$LINES" -lt 100 && "${BINARY_HIT:-0}" -eq 0 ]]; then
     # Binary files have no line count but represent real changes -- treat as standard+
-    TIER=express
+    TIER_RESOLVED=express
   else
-    TIER=standard
+    TIER_RESOLVED=standard
   fi
 fi
+if ! _gate_assurance_policy_lookup tiers tier "$TIER_RESOLVED" default_reviewers >/dev/null; then
+  printf 'Error: detected gate tier is absent from policy: %s\n' "$TIER_RESOLVED" >&2
+  exit 2
+fi
+TIER="$TIER_RESOLVED"
+TIER_EVIDENCE_FLOOR="$(_gate_assurance_policy_lookup tiers tier "$TIER" evidence_floor)" || {
+  printf 'Error: gate tier policy has no evidence floor for: %s\n' "$TIER" >&2
+  exit 2
+}
 
 # ── Brief-based tier suggestion (advisory; never overrides --tier) ────────────
 if [[ -n "$BRIEF_FILE" && -f "$BRIEF_FILE" && -z "$TIER_OVERRIDE" ]]; then
@@ -888,21 +1290,21 @@ if [[ -n "$BRIEF_FILE" && -f "$BRIEF_FILE" && -z "$TIER_OVERRIDE" ]]; then
 fi
 
 # ── Determine reviewer list ───────────────────────────────────────────────────
-ALL_REVIEWERS="critic qa-tester architecture-reviewer security-reviewer risk-reviewer"
-
 if [[ -n "$REVIEWERS_OVERRIDE" ]]; then
-  REVIEWERS=$(printf '%s' "$REVIEWERS_OVERRIDE" | tr ',' ' ')
+  REVIEWERS="$(_gate_normalize_reviewer_list "$REVIEWERS_OVERRIDE" "$REVIEWERS_OPTION_SOURCE")" || exit 2
+  COVERAGE_REQUESTED_DISPLAY="$(printf '%s' "$REVIEWERS" | tr ' ' ',')"
 else
-  case "$TIER" in
-    express)  REVIEWERS="critic qa-tester";;
-    standard) REVIEWERS="critic qa-tester architecture-reviewer";;
-    full)     REVIEWERS="$ALL_REVIEWERS";;
-    *)        REVIEWERS="$ALL_REVIEWERS";;
-  esac
+  _tier_default_reviewers="$(_gate_assurance_policy_lookup tiers tier "$TIER" default_reviewers)" || {
+    printf 'Error: gate tier policy has no default reviewer coverage for: %s\n' "$TIER" >&2
+    exit 2
+  }
+  REVIEWERS="$(_gate_normalize_reviewer_list "$_tier_default_reviewers" "tier $TIER default_reviewers")" || exit 2
+  COVERAGE_REQUESTED_DISPLAY="default"
+  unset _tier_default_reviewers
 fi
 
 REVIEWER_DISPLAY=$(printf '%s' "$REVIEWERS" | tr ' ' ',')
-NUM_REVIEWERS=$(printf '%s\n' $REVIEWERS | wc -l | tr -d ' ')
+NUM_REVIEWERS=$(printf '%s\n' "$REVIEWERS" | awk '{print NF}')
 
 # Compute skipped dimensions
 SKIPPED=""
@@ -912,6 +1314,8 @@ for r in $ALL_REVIEWERS; do
   fi
 done
 SKIPPED_DISPLAY="${SKIPPED:-none}"
+COVERAGE_SELECTED_DISPLAY="$REVIEWER_DISPLAY"
+COVERAGE_SKIPPED_DISPLAY="$SKIPPED_DISPLAY"
 
 # ── Resolve reviewer definitions ─────────────────────────────────────────────
 # Definitions outside the reviewed workspace are trusted installation assets.
@@ -979,6 +1383,16 @@ OUTPUT_FILE="${OUTPUT_OVERRIDE:-$WORK_DIR/.gate-results/gate-${TIMESTAMP}.md}"
 # here IS the absolute working dir and a relative OUTPUT_FILE resolves to an absolute path under it.
 [[ "$OUTPUT_FILE" = /* ]] || OUTPUT_FILE="$PWD/$OUTPUT_FILE"
 mkdir -p "$(dirname "$OUTPUT_FILE")"
+_output_parent="$(cd "$(dirname "$OUTPUT_FILE")" && pwd -P)"
+OUTPUT_FILE="$_output_parent/$(basename "$OUTPUT_FILE")"
+unset _output_parent
+if [[ -n "$INITIAL_RESULT_RESOLVED" \
+    && ( "$OUTPUT_FILE" == "$INITIAL_RESULT_RESOLVED" \
+      || ( -e "$OUTPUT_FILE" && "$OUTPUT_FILE" -ef "$INITIAL_RESULT_RESOLVED" ) ) ]]; then
+  printf 'Error: --output must not overwrite the referenced --initial-result: %s\n' \
+    "$INITIAL_RESULT_RESOLVED" >&2
+  exit 2
+fi
 touch "$OUTPUT_FILE"
 
 # Track all brief files for EXIT cleanup
@@ -1131,7 +1545,22 @@ render_gate_overrides_block() {
 # Pre-format the override block for heredoc injection (empty when no overrides).
 GATE_OVERRIDES_CONTEXT_BLOCK="$(render_gate_overrides_block "$GATE_OVERRIDES_CONTENT")"
 
-say 'pr-gate: %s tier -- %s\n' "$TIER" "$REVIEWER_DISPLAY"
+INITIAL_RESULT_DISPLAY="${INITIAL_RESULT_RESOLVED:-none}"
+printf -v GATE_ASSURANCE_CONTEXT_BLOCK \
+  '  Assurance coordinates (resolved by the gate shell; do not reinterpret):\n    tier.requested: %s\n    tier.resolved: %s\n    tier.evidence_floor: %s\n    mode.requested: %s\n    mode.resolved: %s\n    mode.topology: %s\n    mode.synthesis: %s\n    pass.requested: %s\n    pass.resolved: %s\n    pass.scope: %s\n    pass.initial_result: %s\n    coverage.requested: %s\n    coverage.selected: %s\n    coverage.skipped: %s\n    policy.source: %s\n' \
+  "$TIER_REQUESTED" "$TIER_RESOLVED" "$TIER_EVIDENCE_FLOOR" \
+  "$MODE_REQUESTED" "$MODE_RESOLVED" "$MODE_TOPOLOGY" "$MODE_SYNTHESIS" \
+  "$PASS_KIND_REQUESTED" "$PASS_KIND_RESOLVED" "$PASS_SCOPE" "$INITIAL_RESULT_DISPLAY" \
+  "$COVERAGE_REQUESTED_DISPLAY" \
+  "$COVERAGE_SELECTED_DISPLAY" "$COVERAGE_SKIPPED_DISPLAY" \
+  "$GATE_ASSURANCE_POLICY_SOURCE"
+
+say 'pr-gate: tier %s -> %s; mode %s -> %s; pass %s -> %s\n' \
+  "$TIER_REQUESTED" "$TIER_RESOLVED" "$MODE_REQUESTED" "$MODE_RESOLVED" \
+  "$PASS_KIND_REQUESTED" "$PASS_KIND_RESOLVED"
+say 'pr-gate: coverage requested=%s selected=%s skipped=%s; policy=%s\n' \
+  "$COVERAGE_REQUESTED_DISPLAY" "$COVERAGE_SELECTED_DISPLAY" \
+  "$COVERAGE_SKIPPED_DISPLAY" "$GATE_ASSURANCE_POLICY_SOURCE"
 [[ "${ADJ_COUNT:-0}" -gt 0 ]] && say '  adjacent test files added: %d\n' "$ADJ_COUNT"
 say 'result will be written to: %s\n\n' "$OUTPUT_FILE"
 
@@ -1589,7 +2018,7 @@ context:
   Base: ${BASE}${HEAD_METADATA_LINE}
   Scope: ${SCOPE:-none}
   Date: $(date '+%Y-%m-%d')
-${GATE_OVERRIDES_CONTEXT_BLOCK}${TEST_EVIDENCE_CONTEXT_BLOCK}
+${GATE_ASSURANCE_CONTEXT_BLOCK}${GATE_OVERRIDES_CONTEXT_BLOCK}${TEST_EVIDENCE_CONTEXT_BLOCK}
 ${MEMORY_CONTEXT_BLOCK}
   Verified reference files (exist in working tree -- check before citing):
 ${REPO_REF_INDEX}
@@ -1632,8 +2061,8 @@ output_format: |
   ---
   gate_result_version: pr_gate_result_v1
   final: GO|NO-GO
-  tier: express|standard|full|targeted
-  mode: sequential
+  tier: ${TIER}
+  mode: ${MODE_RESOLVED}
   most_severe: approve|advise|block-soft|block
   reviewers:
     critic: approve|advise|block-soft|skipped
@@ -1823,7 +2252,7 @@ context:
   Base: ${BASE}${HEAD_METADATA_LINE}
   Scope: ${SCOPE:-none}
   Date: $(date '+%Y-%m-%d')
-${GATE_OVERRIDES_CONTEXT_BLOCK}${TEST_EVIDENCE_CONTEXT_BLOCK}
+${GATE_ASSURANCE_CONTEXT_BLOCK}${GATE_OVERRIDES_CONTEXT_BLOCK}${TEST_EVIDENCE_CONTEXT_BLOCK}
 ${MEMORY_CONTEXT_BLOCK}
   Verified reference files (exist in working tree -- check before citing):
 ${REPO_REF_INDEX}
@@ -2045,7 +2474,7 @@ context:
   Base: ${BASE}${HEAD_METADATA_LINE}
   Scope: ${SCOPE:-none}
   Date: $(date '+%Y-%m-%d')
-${GATE_OVERRIDES_CONTEXT_BLOCK}${TEST_EVIDENCE_CONTEXT_BLOCK}
+${GATE_ASSURANCE_CONTEXT_BLOCK}${GATE_OVERRIDES_CONTEXT_BLOCK}${TEST_EVIDENCE_CONTEXT_BLOCK}
   Verified reference files (exist in working tree -- check before citing):
 ${REPO_REF_INDEX}
   Reviewer findings (embedded -- do NOT attempt to read any external reviewer output file):
@@ -2081,7 +2510,7 @@ output_format: |
   gate_result_version: pr_gate_result_v1
   final: GO|NO-GO
   tier: ${TIER}
-  mode: parallel
+  mode: ${MODE_RESOLVED}
   most_severe: approve|advise|block-soft|block
   reviewers:
     critic: approve|advise|block-soft|skipped
diff --git a/skills/pr-gate-review/SKILL.md b/skills/pr-gate-review/SKILL.md
index feb20d82..6514fa66 100644
--- a/skills/pr-gate-review/SKILL.md
+++ b/skills/pr-gate-review/SKILL.md
@@ -18,16 +18,16 @@ implement → pr-gate → fix NO-GO → push → PR).
 
 - Slash command: `/pr-gate` (see `commands/pr-gate.md`). It dispatches the
   reviewers and writes a typed result to `.gate-results/`.
-- Direct: `bash runtime/bin/pr-gate.sh --cd <repo> --executor auto [--parallel]`.
+- Direct: `bash runtime/bin/pr-gate.sh --cd <repo> --executor auto [--mode sequential|parallel]`.
 - Reasoning effort defaults to `medium` (`--effort low|medium|high`, independent of `--model`/`--executor`). Only reach for `--effort high` when you need deeper analysis — e.g. a hard-to-diagnose finding, or escalating after repeated NO-GO rounds on the same issue.
 
 **Tier / mode** (tiers reflect rigor level, not reviewer count):
 - `express` — hotfix, docs-only, `architecture_impact: none`; machine verify + critic + qa.
 - `standard` — feature, `architecture_impact: minor`; adds architecture-reviewer with conceptual map.
-- `full` — architectural change, `architecture_impact: major`, sensitive path; parallel cross-context + security + risk hard gates.
-- Default = **sequential**, low token cost. `--parallel` gives each reviewer an independent session.
+- `full` — architectural change, `architecture_impact: major`, sensitive path; defaults to all reviewer dimensions.
+- Default = **sequential**, low token cost. `--mode parallel` gives each reviewer an independent session; `--parallel` remains a compatibility spelling.
 - Pass `--brief <file>` to get an advisory tier suggestion based on the brief's `architecture_impact`.
-- Force a tier with `--tier express|standard|full`; re-gate a subset with `--targeted r1,r2`.
+- Force a tier with `--tier express|standard|full`; re-gate a remediation subset with `--targeted r1,r2 --initial-result <path>`.
 
 ## Reading the result
 
diff --git a/tests/bin/run-tests.sh b/tests/bin/run-tests.sh
index aa4eca3b..5f4fc7c4 100755
--- a/tests/bin/run-tests.sh
+++ b/tests/bin/run-tests.sh
@@ -215,6 +215,8 @@ map_path() {
       add_suite test-pmctl-task; behavioral=1 ;;
     core/schema/preflight-evidence.schema.json)
       add_suite test-pr-gate; behavioral=1 ;;
+    core/policy/gate-tiers.tsv|core/policy/gate-modes.tsv|core/policy/gate-pass-kinds.tsv)
+      add_suite test-pr-gate; add_suite test-pr-gate-profile; behavioral=1 ;;
     runtime/lib/pmctl-config.sh)
       add_suite test-pmctl-dispatch; add_suite test-pmctl-memory; add_suite test-pmctl-context; behavioral=1 ;;
     runtime/lib/memory.sh|runtime/lib/memory-dir.sh)
diff --git a/tests/shell/test-commands.sh b/tests/shell/test-commands.sh
index 23a7afe6..b372cfdf 100755
--- a/tests/shell/test-commands.sh
+++ b/tests/shell/test-commands.sh
@@ -477,6 +477,59 @@ STUB
 }
 pr_gate_fence_executor_model_behavior
 
+pr_gate_fence_assurance_behavior() {
+  local name="pr-gate: assurance argument parser behavior"
+  should_run "$name" || return 0
+  local case_dir="$tmp_root/pr-gate-fence-assurance"
+  local snippet="$case_dir/snippet.sh" bin_dir="$case_dir/bin"
+  local args_log="$case_dir/args.log" out="$case_dir/out" err="$case_dir/err"
+  local valid_status missing_initial_status initial_only_status conflict_status invalid_mode_status
+  mkdir -p "$bin_dir"
+  extract_pr_gate_bash_fence "$snippet"
+  cat > "$bin_dir/pmctl" <<'STUB'
+#!/usr/bin/env bash
+printf '%s\n' "$@" > "${PR_GATE_ARGS_LOG:?}"
+STUB
+  chmod +x "$bin_dir/pmctl"
+
+  ARGUMENTS="--executor claude --targeted critic --initial-result /tmp/initial.md --mode parallel" \
+    PR_GATE_ARGS_LOG="$args_log" PATH="$bin_dir:$PATH" bash "$snippet" > "$out" 2> "$err"
+  valid_status=$?
+  if [[ "$valid_status" -ne 0 ]] ||
+     ! grep -qx -- '--targeted' "$args_log" ||
+     ! grep -qx 'critic' "$args_log" ||
+     ! grep -qx -- '--initial-result' "$args_log" ||
+     ! grep -qx '/tmp/initial.md' "$args_log" ||
+     ! grep -qx -- '--mode' "$args_log" ||
+     ! grep -qx 'parallel' "$args_log" ||
+     grep -qx -- '--reviewers' "$args_log"; then
+    fail "$name" "valid targeted invocation was not forwarded canonically"
+    return
+  fi
+
+  rm -f "$args_log"
+  ARGUMENTS="--executor claude --targeted critic" \
+    PR_GATE_ARGS_LOG="$args_log" PATH="$bin_dir:$PATH" bash "$snippet" > "$out" 2> "$err"
+  missing_initial_status=$?
+  ARGUMENTS="--executor claude --initial-result /tmp/initial.md" \
+    PR_GATE_ARGS_LOG="$args_log" PATH="$bin_dir:$PATH" bash "$snippet" > "$out" 2> "$err"
+  initial_only_status=$?
+  ARGUMENTS="--executor claude --mode parallel --sequential" \
+    PR_GATE_ARGS_LOG="$args_log" PATH="$bin_dir:$PATH" bash "$snippet" > "$out" 2> "$err"
+  conflict_status=$?
+  ARGUMENTS="--executor claude --mode default" \
+    PR_GATE_ARGS_LOG="$args_log" PATH="$bin_dir:$PATH" bash "$snippet" > "$out" 2> "$err"
+  invalid_mode_status=$?
+
+  if [[ "$missing_initial_status" -eq 2 && "$initial_only_status" -eq 2 &&
+        "$conflict_status" -eq 2 && "$invalid_mode_status" -eq 2 ]]; then
+    pass "$name"
+  else
+    fail "$name" "statuses missing_initial=$missing_initial_status initial_only=$initial_only_status conflict=$conflict_status invalid_mode=$invalid_mode_status"
+  fi
+}
+pr_gate_fence_assurance_behavior
+
 assert_frontmatter "ship: frontmatter valid" "$SHIP"
 should_run "ship: scoped to a single named ticket per invocation" && assert_file_contains "ship: scoped to a single named ticket per invocation" "$SHIP" "one ticket per invocation" && pass "ship: scoped to a single named ticket per invocation"
 should_run "ship: does not batch-scan BACKLOG for candidates" && assert_file_contains "ship: does not batch-scan BACKLOG for candidates" "$SHIP" "Do not scan" && pass "ship: does not batch-scan BACKLOG for candidates"
@@ -536,7 +589,7 @@ fi
 should_run "ship: explains why detached+wait is unnecessary here" && assert_file_contains "ship: explains why detached+wait is unnecessary here" "$SHIP" "nothing else for the main thread to do while it waits" && pass "ship: explains why detached+wait is unnecessary here"
 should_run "ship: reads Final GO/NO-GO verdict" && assert_file_contains "ship: reads Final GO/NO-GO verdict" "$SHIP" "Final:" && pass "ship: reads Final GO/NO-GO verdict"
 should_run "ship: NO-GO fixes every finding not only blocking ones" && assert_file_contains "ship: NO-GO fixes every finding not only blocking ones" "$SHIP" "the blocking ones" && pass "ship: NO-GO fixes every finding not only blocking ones"
-should_run "ship: re-runs gate with --reviewers targeting" && assert_file_contains "ship: re-runs gate with --reviewers targeting" "$SHIP" "--reviewers <reviewer,...>" && pass "ship: re-runs gate with --reviewers targeting"
+should_run "ship: re-runs gate with explicit targeted initial reference" && assert_file_contains "ship: re-runs gate with explicit targeted initial reference" "$SHIP" "--targeted <reviewer,...>" && assert_file_contains "ship: re-runs gate with explicit targeted initial reference" "$SHIP" "--initial-result" && pass "ship: re-runs gate with explicit targeted initial reference"
 should_run "ship: references project-pm Rules A/B synthesis" && assert_file_contains "ship: references project-pm Rules A/B synthesis" "$SHIP" "Rules A/B" && pass "ship: references project-pm Rules A/B synthesis"
 should_run "project-pm: classifies refactor/reuse recheck after gate fixes" && assert_file_contains "project-pm: classifies refactor/reuse recheck after gate fixes" "$PROJECT_PM" "refactor_reuse_recheck: required" && assert_file_contains "project-pm: classifies refactor/reuse recheck after gate fixes" "$PROJECT_PM" "refactor_reuse_recheck: skip" && pass "project-pm: classifies refactor/reuse recheck after gate fixes"
 # exactly two stop conditions, no more
diff --git a/tests/shell/test-core-schemas.sh b/tests/shell/test-core-schemas.sh
index 01ffd81a..0c6be042 100755
--- a/tests/shell/test-core-schemas.sh
+++ b/tests/shell/test-core-schemas.sh
@@ -3,13 +3,13 @@
 #
 # These tests validate that:
 #   1. Every JSON Schema file under core/schema/ is valid JSON
-#   2. Every YAML file under core/policy/ and core/state/ is valid YAML
+#   2. Every YAML/TSV file under core/policy/ and core/state/ has valid structure
 #   3. Schemas declaring schema_version do so as a positive integer `const`
 #   4. Enum values referenced inline in schemas stay in sync with the
-#      corresponding policy YAML files (the documented editing source).
+#      corresponding declarative policy files (the documented editing source).
 #
 # Per docs/spikes/CC-229-substrate-synthesis.md §E Q1: JSON Schema is the
-# ajv-compliant source-of-truth; policy YAML is the human editing surface.
+# ajv-compliant source-of-truth; policy YAML/TSV is the human editing surface.
 # This test enforces the sync that comment-level documentation requests.
 #
 # Runs via: tests/shell/test-core-schemas.sh
@@ -60,6 +60,29 @@ _schema_enum() {
   jq -r "$path | .[]" "$file" 2>/dev/null || true
 }
 
+_tsv_column() {
+  # Extract a named column from a headered TSV, excluding comments/blank lines.
+  local file="$1" column="$2"
+  awk -F '\t' -v wanted="$column" '
+    /^[[:space:]]*#/ || /^[[:space:]]*$/ { next }
+    !header_seen {
+      header_seen=1
+      width=NF
+      for (i=1; i<=NF; i++) {
+        sub(/\r$/, "", $i)
+        if ($i == wanted) column_index=i
+      }
+      if (!column_index) exit 2
+      next
+    }
+    {
+      sub(/\r$/, "", $NF)
+      if (NF != width || $(column_index) == "") exit 2
+      print $(column_index)
+    }
+  ' "$file"
+}
+
 # ---------- tests ----------
 
 case_schema_parse() {
@@ -103,6 +126,35 @@ case_yaml_parse() {
   pass "$name"
 }
 
+case_tsv_parse() {
+  # Verifies that a policy TSV has a header, at least one row, consistent
+  # tab-separated width, and no empty cells.
+  local file="$1"
+  local name="TSV: $file has valid structure"
+  should_run "$name" || return 0
+  if awk -F '\t' '
+      /^[[:space:]]*#/ || /^[[:space:]]*$/ { next }
+      !header_seen {
+        header_seen=1
+        width=NF
+        for (i=1; i<=NF; i++) if ($i == "") exit 2
+        next
+      }
+      {
+        rows++
+        if (NF != width) exit 2
+        for (i=1; i<=NF; i++) if ($i == "") exit 2
+      }
+      END {
+        if (!header_seen || width < 2 || rows < 1) exit 2
+      }
+    ' "$file"; then
+    pass "$name"
+  else
+    fail "$name" "invalid header or row structure"
+  fi
+}
+
 case_schema_version_const() {
   # Verifies schema_version is a positive integer const OR an enum of positive integers.
   # context-pack uses enum [1,2] for v1/v2 compat; all others use const.
@@ -153,6 +205,28 @@ case_enum_sync() {
   fi
 }
 
+case_review_tier_policy_compatibility() {
+  # Verify current Review tiers match the canonical TSV while schema_version 1
+  # retains only `targeted` as its documented legacy compatibility value.
+  local schema_file="$CORE_DIR/schema/review.schema.json"
+  local tsv_file="$CORE_DIR/policy/gate-tiers.tsv"
+  local name="enum-sync: review tiers == gate-tiers.tsv plus legacy targeted"
+  should_run "$name" || return 0
+
+  local current_schema_vals tsv_vals legacy_count
+  current_schema_vals=$(_schema_enum "$schema_file" '.properties.tier.enum' \
+    | grep -vx 'targeted' | sort)
+  tsv_vals=$(_tsv_column "$tsv_file" "tier" | sort)
+  legacy_count=$(_schema_enum "$schema_file" '.properties.tier.enum' \
+    | grep -cx 'targeted' || true)
+
+  if [[ "$current_schema_vals" == "$tsv_vals" && "$legacy_count" -eq 1 ]]; then
+    pass "$name"
+  else
+    fail "$name" "current tiers differ from TSV or legacy targeted is not unique"
+  fi
+}
+
 # ---------- run ----------
 
 # 1. JSON Schema parse
@@ -165,6 +239,11 @@ for f in "$CORE_DIR"/policy/*.yaml "$CORE_DIR"/state/layout.yaml; do
   case_yaml_parse "$f"
 done
 
+# 2b. TSV parse
+for f in "$CORE_DIR"/policy/*.tsv; do
+  case_tsv_parse "$f"
+done
+
 # 3. schema_version: const 1 on every payload schema that declares it
 for f in "$CORE_DIR"/schema/*.schema.json; do
   # Skip schemas that intentionally omit schema_version (none in M1)
@@ -222,6 +301,10 @@ case_enum_sync "$CORE_DIR/schema/review.schema.json" \
   "$CORE_DIR/policy/reviewer-policy.yaml" \
   "verdicts"
 
+# Gate rigor tiers: new producers use the TSV; Review schema v1 still accepts
+# its historical targeted value so old state remains structurally valid.
+case_review_tier_policy_compatibility
+
 case_enum_sync_crlf_input() {
   # A YAML file checked out with CRLF endings (Windows core.autocrlf re-applying
   # despite .gitattributes, or a non-git copy) must not produce a false enum-sync
diff --git a/tests/shell/test-layer-boundaries.sh b/tests/shell/test-layer-boundaries.sh
index a3a7fe35..897e4baf 100755
--- a/tests/shell/test-layer-boundaries.sh
+++ b/tests/shell/test-layer-boundaries.sh
@@ -15,7 +15,7 @@
 #
 # Boundary the rules encode (calibrated against the real tree, CC-233):
 #   - `core/` is definitions only: no shell/executables, no CLI-product-named
-#     files/dirs, only .yaml/.json/.md, no CLI product name as a field-name KEY.
+#     files/dirs, only .yaml/.json/.md/.tsv, no CLI product name as a field-name KEY.
 #     (CLI names MAY appear as enum VALUES / schema descriptions / prose — those
 #      are data, not structure; the rules check structure, ignore prose.)
 #   - `adapters/**/*.sh` must NOT call the shared FLOW (brief-validate, guard,
@@ -55,12 +55,13 @@ check_core_no_cli_named_paths() {
     -o -iname '*antigravity*' -o -iname '*opencode*' \) 2>/dev/null
 }
 
-# C3: core/ holds only declarative file types (.yaml/.yml/.json/.md).
+# C3: core/ holds only declarative file types (.yaml/.yml/.json/.md/.tsv).
 check_core_only_declarative() {
   local root="$1"
   [[ -d "$root/core" ]] || return 0
   find "$root/core" -type f \
-    ! \( -name '*.yaml' -o -name '*.yml' -o -name '*.json' -o -name '*.md' \) 2>/dev/null
+    ! \( -name '*.yaml' -o -name '*.yml' -o -name '*.json' -o -name '*.md' \
+      -o -name '*.tsv' \) 2>/dev/null
 }
 
 # C4: no CLI product name appears as a field-name KEY (YAML/JSON) in core/.
diff --git a/tests/shell/test-pr-gate.sh b/tests/shell/test-pr-gate.sh
index 7bf37a9a..811e0cc8 100755
--- a/tests/shell/test-pr-gate.sh
+++ b/tests/shell/test-pr-gate.sh
@@ -68,6 +68,9 @@ create_runner() {
   cp -R "$REPO_ROOT/runtime/lib/." "$dir/lib/"
   mkdir -p "$dir/core/policy"
   cp "$REPO_ROOT/core/policy/isolation-level.yaml" "$dir/core/policy/isolation-level.yaml"
+  cp "$REPO_ROOT/core/policy/gate-tiers.tsv" "$dir/core/policy/gate-tiers.tsv"
+  cp "$REPO_ROOT/core/policy/gate-modes.tsv" "$dir/core/policy/gate-modes.tsv"
+  cp "$REPO_ROOT/core/policy/gate-pass-kinds.tsv" "$dir/core/policy/gate-pass-kinds.tsv"
   mkdir -p "$dir/adapters/codex"
   cat > "$dir/adapters/codex/dispatch.sh" <<'STUB_EOF'
 #!/usr/bin/env bash
@@ -503,6 +506,135 @@ run_gate() {
   return "$code"
 }
 
+write_valid_initial_gate_result() {
+  local path="$1" final="${2:-NO-GO}"
+  cat > "$path" << INITIAL_GATE_EOF
+---
+gate_result_version: pr_gate_result_v1
+final: ${final}
+tier: standard
+mode: sequential
+most_severe: block
+reviewers:
+  critic: block
+escalation:
+  recommended: false
+  reviewers: []
+  reason: []
+---
+
+# Initial gate result
+
+## Gate Conclusion
+Final: ${final}
+INITIAL_GATE_EOF
+}
+
+# Behavior: the bounded copy-mode policy snapshot is byte-for-byte equivalent
+# to all three canonical gate policy TSV sources.
+# Steps: extract each generated heredoc from pr-gate.sh, compare it with the
+# matching core/policy file, and fail on any drift.
+test_gate_assurance_policy_snapshot_matches_sources() {
+  local name="gate-assurance-policy-snapshot-matches-sources"
+  should_run "$name" || return 0
+  local table delimiter source snapshot
+  for table in tiers modes pass-kinds; do
+    case "$table" in
+      tiers)
+        delimiter="GATE_ASSURANCE_TIERS_TSV"
+        source="$REPO_ROOT/core/policy/gate-tiers.tsv"
+        ;;
+      modes)
+        delimiter="GATE_ASSURANCE_MODES_TSV"
+        source="$REPO_ROOT/core/policy/gate-modes.tsv"
+        ;;
+      pass-kinds)
+        delimiter="GATE_ASSURANCE_PASS_KINDS_TSV"
+        source="$REPO_ROOT/core/policy/gate-pass-kinds.tsv"
+        ;;
+    esac
+    snapshot="$(awk -v marker="$delimiter" '
+      index($0, "cat <<\047" marker "\047") { inside=1; next }
+      inside && $0 == marker { exit }
+      inside { print }
+    ' "$REPO_ROOT/runtime/bin/pr-gate.sh")"
+    if [[ "$snapshot" != "$(cat "$source")" ]]; then
+      fail "$name" "generated snapshot drifted from $source"
+      return
+    fi
+  done
+  pass "$name"
+}
+
+# Behavior: a repo-layout/copy-bundle policy source controls tier-default
+# reviewer selection instead of the generated fallback or a hardcoded case.
+# Steps: change only the copied express default to critic, run a docs gate, and
+# assert the captured brief selects critic while retaining express tier.
+test_gate_tier_policy_source_controls_default_coverage() {
+  local name="gate-tier-policy-source-controls-default-coverage"
+  should_run "$name" || return 0
+  local dir="$TMP_ROOT/$name"
+  local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
+  local out="$dir/out" err="$dir/err" brief="$dir/brief.md" rewritten="$dir/gate-tiers.tsv"
+  mkdir -p "$dir"
+  create_runner "$runner"
+  create_agents "$home" critic qa-tester architecture-reviewer security-reviewer risk-reviewer
+  create_repo "$repo" docs
+  awk -F '\t' -v OFS='\t' '
+    $1 == "express" { $2="critic" }
+    { print }
+  ' "$runner/core/policy/gate-tiers.tsv" > "$rewritten"
+  mv "$rewritten" "$runner/core/policy/gate-tiers.tsv"
+
+  set +e
+  CODEX_GATE_CAPTURE_BRIEF="$brief" run_gate "$home" "$runner" "$repo" "$out" "$err" --base main
+  local code=$?
+  set -e
+  if [[ "$code" -ne 0 ]]; then
+    fail "$name" "exit $code, expected 0"
+    return
+  fi
+  assert_file_contains "$name" "$brief" "Tier: express" || return
+  assert_file_contains "$name" "$brief" "coverage.selected: critic" || return
+  assert_file_contains "$name" "$brief" "Reviewers: critic" || return
+  assert_not_contains "$name" "$brief" "Process each reviewer IN ORDER: critic,qa-tester" || return
+  pass "$name"
+}
+
+# Behavior: a copied gate without canonical policy files resolves the same
+# defaults from its bounded generated snapshot and reports the degraded source.
+# Steps: remove the copied TSV files, run a docs gate, and assert express /
+# sequential / initial defaults plus generated-snapshot provenance.
+test_gate_assurance_policy_snapshot_is_copy_mode_fallback() {
+  local name="gate-assurance-policy-snapshot-is-copy-mode-fallback"
+  should_run "$name" || return 0
+  local dir="$TMP_ROOT/$name"
+  local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
+  local out="$dir/out" err="$dir/err" brief="$dir/brief.md"
+  mkdir -p "$dir"
+  create_runner "$runner"
+  create_agents "$home" critic qa-tester architecture-reviewer security-reviewer risk-reviewer
+  create_repo "$repo" docs
+  rm -f \
+    "$runner/core/policy/gate-tiers.tsv" \
+    "$runner/core/policy/gate-modes.tsv" \
+    "$runner/core/policy/gate-pass-kinds.tsv"
+
+  set +e
+  CODEX_GATE_CAPTURE_BRIEF="$brief" run_gate "$home" "$runner" "$repo" "$out" "$err" --base main
+  local code=$?
+  set -e
+  if [[ "$code" -ne 0 ]]; then
+    fail "$name" "exit $code, expected 0"
+    return
+  fi
+  assert_file_contains "$name" "$brief" "tier.resolved: express" || return
+  assert_file_contains "$name" "$brief" "mode.resolved: sequential" || return
+  assert_file_contains "$name" "$brief" "pass.resolved: initial" || return
+  assert_file_contains "$name" "$brief" "policy.source: generated-snapshot" || return
+  pass "$name"
+}
+
 # Behavior: express-tier diff with no overrides routes to codex with the
 # express reviewer set (critic, qa-tester).
 # Steps: run the gate on a docs-only diff, assert stderr shows dispatch
@@ -530,6 +662,15 @@ test_tier_detection() {
   # Sequential dispatch chatter now lands on stderr, not stdout.
   assert_file_contains "$name" "$err" "DISPATCH_STUB:success" || return
   assert_file_contains "$name" "$brief" "Tier: express" || return
+  assert_file_contains "$name" "$brief" "tier.requested: auto" || return
+  assert_file_contains "$name" "$brief" "tier.resolved: express" || return
+  assert_file_contains "$name" "$brief" "tier.evidence_floor: reviewer-verdicts" || return
+  assert_file_contains "$name" "$brief" "mode.requested: default" || return
+  assert_file_contains "$name" "$brief" "mode.resolved: sequential" || return
+  assert_file_contains "$name" "$brief" "mode.synthesis: inline" || return
+  assert_file_contains "$name" "$brief" "pass.resolved: initial" || return
+  assert_file_contains "$name" "$brief" "coverage.requested: default" || return
+  assert_file_contains "$name" "$brief" "coverage.selected: critic,qa-tester" || return
   assert_file_contains "$name" "$brief" "Executor: codex" || return
   assert_file_contains "$name" "$brief" "Reviewers: critic,qa-tester" || return
   pass "$name"
@@ -955,14 +1096,13 @@ test_no_changed_files() {
   pass "$name"
 }
 
-# Behavior: an explicit --reviewers list overrides tier auto-detection
-# (Tier: targeted) and the parallel synthesis brief embeds reviewer findings
-# inline rather than pointing at read: paths for reviewers not in the list.
+# Behavior: an explicit --reviewers list overrides requested coverage without
+# changing auto-detected tier or review pass kind. The parallel synthesis brief
+# embeds only selected reviewer findings inline.
 # Steps: run the gate with --reviewers critic --parallel against a diff that
-# would otherwise tier-detect to something else, and assert the captured
-# synthesis brief has Tier: targeted, Reviewers: critic, inline
-# "--- critic findings ---", and no reviewer-critic-/qa-tester read: paths.
-test_reviewers_override_skips_tier_detection() {
+# tier-detects to standard, and assert standard/initial coordinates, critic-only
+# coverage, inline findings, and no reviewer output read paths.
+test_reviewers_override_preserves_tier_detection() {
   local name="reviewers-override"
   should_run "$name" || return 0
   local dir="$TMP_ROOT/$name"
@@ -971,7 +1111,7 @@ test_reviewers_override_skips_tier_detection() {
   mkdir -p "$dir"
   create_runner "$runner"
   create_agents "$home" critic
-  create_repo "$repo" many
+  create_repo_with_branch "$repo" standard
 
   set +e
   CODEX_GATE_CAPTURE_BRIEF="$brief" run_gate "$home" "$runner" "$repo" "$out" "$err" --base main --reviewers critic --parallel
@@ -982,7 +1122,12 @@ test_reviewers_override_skips_tier_detection() {
     return
   fi
   # Parallel mode: CAPTURE_BRIEF receives the synthesis brief (last dispatch)
-  assert_file_contains "$name" "$brief" "Tier: targeted" || return
+  assert_file_contains "$name" "$brief" "Tier: standard" || return
+  assert_file_contains "$name" "$brief" "tier.requested: auto" || return
+  assert_file_contains "$name" "$brief" "tier.resolved: standard" || return
+  assert_file_contains "$name" "$brief" "pass.resolved: initial" || return
+  assert_file_contains "$name" "$brief" "coverage.requested: critic" || return
+  assert_file_contains "$name" "$brief" "coverage.selected: critic" || return
   assert_file_contains "$name" "$brief" "Executor: codex" || return
   assert_file_contains "$name" "$brief" "Reviewers: critic" || return
   # Synthesis brief embeds reviewer findings inline — no read: paths to reviewer output files
@@ -992,6 +1137,39 @@ test_reviewers_override_skips_tier_detection() {
   pass "$name"
 }
 
+# Behavior: explicit full-tier intent and critic-only requested coverage remain
+# independent; neither value rewrites the other.
+# Steps: run a docs diff with --tier full --reviewers critic, then assert the
+# combined brief records full requested/resolved tier and critic-only coverage.
+test_full_tier_with_critic_only_coverage_is_truthful() {
+  local name="full-tier-with-critic-only-coverage-is-truthful"
+  should_run "$name" || return 0
+  local dir="$TMP_ROOT/$name"
+  local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
+  local out="$dir/out" err="$dir/err" brief="$dir/brief.md"
+  mkdir -p "$dir"
+  create_runner "$runner"
+  create_agents "$home" critic
+  create_repo "$repo" docs
+
+  set +e
+  CODEX_GATE_CAPTURE_BRIEF="$brief" run_gate "$home" "$runner" "$repo" "$out" "$err" \
+    --base main --tier full --reviewers critic --mode sequential
+  local code=$?
+  set -e
+  if [[ "$code" -ne 0 ]]; then
+    fail "$name" "exit $code, expected 0"
+    return
+  fi
+  assert_file_contains "$name" "$brief" "tier.requested: full" || return
+  assert_file_contains "$name" "$brief" "tier.resolved: full" || return
+  assert_file_contains "$name" "$brief" "mode.resolved: sequential" || return
+  assert_file_contains "$name" "$brief" "coverage.requested: critic" || return
+  assert_file_contains "$name" "$brief" "coverage.selected: critic" || return
+  assert_file_contains "$name" "$brief" "Reviewers: critic" || return
+  pass "$name"
+}
+
 # Behavior: the brief handed to the executor exists on disk at the moment of
 # invocation, and any /tmp brief carries the guard-required `brief-gate-`
 # prefix. On the pmctl dispatch transport that brief is the /tmp/brief-gate-*
@@ -2331,15 +2509,16 @@ test_preflight_runs_even_when_qa_tester_not_targeted() {
   should_run "$name" || return 0
   local dir="$TMP_ROOT/$name"
   local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
-  local out="$dir/out" err="$dir/err" result="$dir/result.md"
+  local out="$dir/out" err="$dir/err" result="$dir/result.md" initial="$dir/initial.md"
   mkdir -p "$dir"
   create_runner "$runner"
   create_agents "$home" critic qa-tester architecture-reviewer security-reviewer risk-reviewer
   create_repo "$repo" docs
+  write_valid_initial_gate_result "$initial"
 
   set +e
   run_gate "$home" "$runner" "$repo" "$out" "$err" --base main \
-    --targeted critic --test-cmd "exit 0" --output "$result"
+    --targeted critic --initial-result "$initial" --test-cmd "exit 0" --output "$result"
   local code=$?
   set -e
   if [[ "$code" -ne 0 ]]; then
@@ -3446,6 +3625,9 @@ test_untracked_binary_routes_to_standard() {
   pass "$name"
 }
 
+run_test test_gate_assurance_policy_snapshot_matches_sources
+run_test test_gate_tier_policy_source_controls_default_coverage
+run_test test_gate_assurance_policy_snapshot_is_copy_mode_fallback
 run_test test_tier_detection
 run_test test_pr_gate_does_not_mutate_gitignore
 run_test test_artifact_filter_drops_gate_artifacts
@@ -3458,7 +3640,8 @@ run_test test_copy_mode_artifact_fallback_body_parity
 run_test test_missing_reviewer_agent
 run_test test_invalid_base_ref
 run_test test_no_changed_files
-run_test test_reviewers_override_skips_tier_detection
+run_test test_reviewers_override_preserves_tier_detection
+run_test test_full_tier_with_critic_only_coverage_is_truthful
 run_test test_brief_file_snapshot_exists_at_dispatch
 run_test test_reviewer_definitions_are_workspace_snapshots
 run_test test_brief_cleanup_on_dispatch_failure
@@ -4085,12 +4268,41 @@ test_copy_mode_dispatches_via_adapter() {
   pass "$name"
 }
 
+# Behavior: --help prints only the bounded user-facing usage contract and
+# includes the canonical assurance flags.
+# Steps: invoke a copied gate with --help and assert current flags are present
+# while shell implementation and generated policy internals are absent.
+test_help_output_is_bounded_and_current() {
+  local name="help-output-is-bounded-and-current"
+  should_run "$name" || return 0
+  local dir="$TMP_ROOT/$name"
+  local runner="$dir/runner" out="$dir/out" err="$dir/err"
+  mkdir -p "$dir"
+  create_runner "$runner"
+
+  set +e
+  "$runner/pr-gate.sh" --help > "$out" 2> "$err"
+  local code=$?
+  set -e
+  if [[ "$code" -ne 0 ]]; then
+    fail "$name" "exit $code, expected 0"
+    return
+  fi
+  assert_file_contains "$name" "$out" "Usage:" || return
+  assert_file_contains "$name" "$out" "--mode <mode>" || return
+  assert_file_contains "$name" "$out" "--targeted <list>" || return
+  assert_file_contains "$name" "$out" "--initial-result <f>" || return
+  assert_not_contains "$name" "$out" "_gate_assurance_policy_snapshot" || return
+  assert_not_contains "$name" "$out" "set -euo pipefail" || return
+  pass "$name"
+}
+
 # Behavior: an unrecognized flag exits 2 and prints an actionable
 # accepted-flags list (not just a bare "Unknown arg"), so callers
 # self-correct on first failure.
 # Steps: run the gate with --bogus-flag, and assert exit 2 and stderr
-# contains "Unknown arg: --bogus-flag", "Accepted:", and
-# "--reviewers|--targeted".
+# contains "Unknown arg: --bogus-flag", "Accepted:", "--targeted", and
+# "--initial-result".
 test_unknown_arg_message() {
   local name="unknown-arg-message"
   should_run "$name" || return 0
@@ -4112,30 +4324,31 @@ test_unknown_arg_message() {
   fi
   assert_file_contains "$name" "$err" "Unknown arg: --bogus-flag" || return
   assert_file_contains "$name" "$err" "Accepted:" || return
-  assert_file_contains "$name" "$err" "--reviewers|--targeted" || return
+  assert_file_contains "$name" "$err" "--targeted" || return
+  assert_file_contains "$name" "$err" "--initial-result" || return
   pass "$name"
 }
 
-# Behavior: --targeted is accepted as an alias of --reviewers (the
-# /pr-gate skill and the script's own comments use "targeted"
-# vocabulary). Scoping a parallel gate to critic must launch critic only --
-# same as --reviewers critic.
-# Steps: run the gate with --targeted critic --parallel, and assert exit
-# 0, no "unknown arg" in stderr, "launched critic" in stdout, and no
-# "launched qa-tester" line.
-test_targeted_alias() {
-  local name="targeted-alias"
+# Behavior: --targeted selects a remediation-delta pass, preserves detected
+# tier, accepts canonical --mode, and carries a structurally valid initial
+# result reference into the resolved assurance context.
+# Steps: create a valid initial result, run a critic-only targeted gate with
+# --mode parallel, and assert the pass/mode/tier/reference/coverage coordinates.
+test_targeted_pass_references_initial_result() {
+  local name="targeted-pass-references-initial-result"
   should_run "$name" || return 0
   local dir="$TMP_ROOT/$name"
   local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
-  local out="$dir/out" err="$dir/err"
+  local out="$dir/out" err="$dir/err" brief="$dir/brief.md" initial="$dir/initial.md"
   mkdir -p "$dir"
   create_runner "$runner"
   create_agents "$home" critic qa-tester architecture-reviewer security-reviewer risk-reviewer
   create_repo "$repo" docs
+  write_valid_initial_gate_result "$initial"
 
   set +e
-  run_gate "$home" "$runner" "$repo" "$out" "$err" --base main --targeted critic --parallel
+  CODEX_GATE_CAPTURE_BRIEF="$brief" run_gate "$home" "$runner" "$repo" "$out" "$err" \
+    --base main --targeted critic --initial-result "$initial" --mode parallel
   local code=$?
   set -e
 
@@ -4143,10 +4356,15 @@ test_targeted_alias() {
     fail "$name" "exit $code, expected 0 (stderr: $(head -3 "$err" 2>/dev/null))"
     return
   fi
-  if grep -qi "unknown arg" "$err"; then
-    fail "$name" "--targeted was rejected as an unknown arg"
-    return
-  fi
+  assert_file_contains "$name" "$brief" "tier.resolved: express" || return
+  assert_file_contains "$name" "$brief" "mode.requested: parallel" || return
+  assert_file_contains "$name" "$brief" "mode.resolved: parallel" || return
+  assert_file_contains "$name" "$brief" "mode.topology: per-reviewer-sessions" || return
+  assert_file_contains "$name" "$brief" "mode.synthesis: separate-session" || return
+  assert_file_contains "$name" "$brief" "pass.resolved: targeted" || return
+  assert_file_contains "$name" "$brief" "pass.scope: remediation-delta" || return
+  assert_file_contains "$name" "$brief" "pass.initial_result: $initial" || return
+  assert_file_contains "$name" "$brief" "coverage.selected: critic" || return
   assert_file_contains "$name" "$out" "launched critic" || return
   if grep -q "launched qa-tester" "$out"; then
     fail "$name" "--targeted critic did not scope reviewers — qa-tester was launched"
@@ -4155,6 +4373,165 @@ test_targeted_alias() {
   pass "$name"
 }
 
+# Behavior: a targeted pass without an initial result fails before dispatch.
+# Steps: invoke --targeted critic without --initial-result and assert exit 2,
+# the explicit requirement error, and no dispatch marker.
+test_targeted_requires_initial_result() {
+  local name="targeted-requires-initial-result"
+  should_run "$name" || return 0
+  local dir="$TMP_ROOT/$name"
+  local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
+  local out="$dir/out" err="$dir/err"
+  mkdir -p "$dir"
+  create_runner "$runner"
+  create_repo "$repo" docs
+
+  set +e
+  run_gate "$home" "$runner" "$repo" "$out" "$err" --base main --targeted critic
+  local code=$?
+  set -e
+  if [[ "$code" -ne 2 ]]; then
+    fail "$name" "exit $code, expected 2"
+    return
+  fi
+  assert_file_contains "$name" "$err" "--targeted requires --initial-result <path>" || return
+  assert_not_contains "$name" "$out" "DISPATCH_STUB" || return
+  pass "$name"
+}
+
+# Behavior: a targeted gate cannot reuse its referenced initial result as the
+# output destination, including a lexical alias of the same path.
+# Steps: pass the initial result back through --output using a ./ alias, assert
+# exit 2 before dispatch, and confirm the initial artifact is unchanged.
+test_targeted_output_cannot_overwrite_initial_result() {
+  local name="targeted-output-cannot-overwrite-initial-result"
+  should_run "$name" || return 0
+  local dir="$TMP_ROOT/$name"
+  local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
+  local out="$dir/out" err="$dir/err" initial="$dir/initial.md" before
+  mkdir -p "$dir"
+  create_runner "$runner"
+  create_repo "$repo" docs
+  write_valid_initial_gate_result "$initial"
+  before="$(cat "$initial")"
+
+  set +e
+  run_gate "$home" "$runner" "$repo" "$out" "$err" --base main \
+    --targeted critic --initial-result "$initial" --output "$dir/./initial.md"
+  local code=$?
+  set -e
+  if [[ "$code" -ne 2 ]]; then
+    fail "$name" "exit $code, expected 2"
+    return
+  fi
+  assert_file_contains "$name" "$err" "--output must not overwrite" || return
+  assert_not_contains "$name" "$out" "DISPATCH_STUB" || return
+  if [[ "$(cat "$initial")" != "$before" ]]; then
+    fail "$name" "initial result content changed"
+    return
+  fi
+  pass "$name"
+}
+
+# Behavior: canonical and compatibility mode spellings fail closed when they
+# request different topologies.
+# Steps: combine --mode parallel with --sequential and assert a controlled
+# exit-2 conflict before dispatch.
+test_conflicting_mode_options_are_rejected() {
+  local name="conflicting-mode-options-are-rejected"
+  should_run "$name" || return 0
+  local dir="$TMP_ROOT/$name"
+  local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
+  local out="$dir/out" err="$dir/err"
+  mkdir -p "$dir"
+  create_runner "$runner"
+  create_repo "$repo" docs
+
+  set +e
+  run_gate "$home" "$runner" "$repo" "$out" "$err" --base main --mode parallel --sequential
+  local code=$?
+  set -e
+  if [[ "$code" -ne 2 ]]; then
+    fail "$name" "exit $code, expected 2"
+    return
+  fi
+  assert_file_contains "$name" "$err" "conflicting gate mode options" || return
+  assert_not_contains "$name" "$out" "DISPATCH_STUB" || return
+  pass "$name"
+}
+
+# Behavior: canonical and compatibility mode spellings may be combined when
+# they request the same topology.
+# Steps: run --mode parallel with --parallel, capture the synthesis brief, and
+# assert one successful parallel resolution.
+test_equivalent_mode_spellings_are_accepted() {
+  local name="equivalent-mode-spellings-are-accepted"
+  should_run "$name" || return 0
+  local dir="$TMP_ROOT/$name"
+  local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
+  local out="$dir/out" err="$dir/err" brief="$dir/brief.md"
+  mkdir -p "$dir"
+  create_runner "$runner"
+  create_agents "$home" critic
+  create_repo "$repo" docs
+
+  set +e
+  CODEX_GATE_CAPTURE_BRIEF="$brief" run_gate "$home" "$runner" "$repo" "$out" "$err" \
+    --base main --reviewers critic --mode parallel --parallel
+  local code=$?
+  set -e
+  if [[ "$code" -ne 0 ]]; then
+    fail "$name" "exit $code, expected 0"
+    return
+  fi
+  assert_file_contains "$name" "$brief" "mode.requested: parallel" || return
+  assert_file_contains "$name" "$brief" "mode.resolved: parallel" || return
+  assert_file_contains "$name" "$out" "launched critic" || return
+  pass "$name"
+}
+
+# Behavior: invalid tier, mode, reviewer, and mixed coverage/pass selectors are
+# rejected as closed CLI inputs instead of falling back to another profile.
+# Steps: run a malformed-input matrix and require exit 2 plus no dispatch for
+# every case.
+test_invalid_assurance_inputs_are_rejected() {
+  local name="invalid-assurance-inputs-are-rejected"
+  should_run "$name" || return 0
+  local dir="$TMP_ROOT/$name"
+  local home="$dir/home" repo="$dir/repo" runner="$dir/runner"
+  local out="$dir/out" err="$dir/err" code args
+  mkdir -p "$dir"
+  create_runner "$runner"
+  create_repo "$repo" docs
+
+  for args in \
+    "--tier targeted" \
+    "--mode default" \
+    "--reviewers unknown-reviewer" \
+    "--reviewers critic,critic" \
+    "--reviewers critic,,qa-tester" \
+    "--initial-result missing.md" \
+    "--reviewers critic --targeted critic"
+  do
+    : > "$out"
+    : > "$err"
+    set +e
+    # shellcheck disable=SC2086 # fixture intentionally expands a small argv matrix
+    run_gate "$home" "$runner" "$repo" "$out" "$err" --base main $args
+    code=$?
+    set -e
+    if [[ "$code" -ne 2 ]]; then
+      fail "$name" "args '$args' exited $code, expected 2"
+      return
+    fi
+    if grep -q "DISPATCH_STUB" "$out" "$err"; then
+      fail "$name" "args '$args' reached dispatch"
+      return
+    fi
+  done
+  pass "$name"
+}
+
 # Behavior: the sequential brief emitted by pr-gate.sh uses ASCII --
 # separators and contains no em dash (U+2014) bytes.
 # Steps: run the gate with --sequential, and assert the captured brief
@@ -4763,8 +5140,14 @@ run_test test_isolation_forwarding_through_pr_gate
 run_test test_effort_forwarding_through_pr_gate
 run_test test_effort_invalid_value_rejected
 run_test test_copy_mode_dispatches_via_adapter
+run_test test_help_output_is_bounded_and_current
 run_test test_unknown_arg_message
-run_test test_targeted_alias
+run_test test_targeted_pass_references_initial_result
+run_test test_targeted_requires_initial_result
+run_test test_targeted_output_cannot_overwrite_initial_result
+run_test test_conflicting_mode_options_are_rejected
+run_test test_equivalent_mode_spellings_are_accepted
+run_test test_invalid_assurance_inputs_are_rejected
 run_test test_seq_brief_ascii_separator
 run_test test_parallel_synthesis_brief_ascii_separator
 run_test test_parallel_reviewer_brief_ascii_separator
diff --git a/tests/shell/test-run-tests.sh b/tests/shell/test-run-tests.sh
index 84a7149b..a47cb064 100755
--- a/tests/shell/test-run-tests.sh
+++ b/tests/shell/test-run-tests.sh
@@ -18,17 +18,20 @@ make_fixture() {
   local name="$1" repo
   repo="$TMP_ROOT/$name"
   mkdir -p "$repo/scripts/lib" "$repo/runtime/lib" "$repo/tests/bin" "$repo/tests/lib" \
-    "$repo/core/schema" "$repo/core/state"
+    "$repo/core/schema" "$repo/core/state" "$repo/core/policy"
   cp "$REPO_ROOT/tests/bin/run-tests.sh" "$repo/tests/bin/run-tests.sh"
   cp "$REPO_ROOT/tests/bin/run-all-tests.sh" "$repo/tests/bin/run-all-tests.sh"
   cp "$SCRIPT_DIR/../lib/test-result.sh" "$repo/tests/lib/test-result.sh"
   cp "$REPO_ROOT/runtime/lib/artifact-paths.sh" "$repo/runtime/lib/artifact-paths.sh"
   cp "$REPO_ROOT/core/schema/test-result.schema.json" "$repo/core/schema/test-result.schema.json"
+  cp "$REPO_ROOT/core/policy/gate-tiers.tsv" "$repo/core/policy/gate-tiers.tsv"
+  cp "$REPO_ROOT/core/policy/gate-modes.tsv" "$repo/core/policy/gate-modes.tsv"
+  cp "$REPO_ROOT/core/policy/gate-pass-kinds.tsv" "$repo/core/policy/gate-pass-kinds.tsv"
   chmod +x "$repo/tests/bin/run-tests.sh"
   cat > "$repo/tests/lib/test-suite-runner.sh" <<'RUNNER'
 #!/usr/bin/env bash
 set -euo pipefail
-suites=(lint-agents lint-scripts lint-script-domain-inventory lint-portable-repo-paths test-lint-shellcheck test-script-domain-inventory test-lint-portable-repo-paths test-lint-frontmatter test-commands test-check-docs-freshness test-guards test-migrate test-install test-doctor test-pmctl-dispatch test-pmctl-context test-pmctl-memory test-pr-gate test-pmctl-operation test-host-manifest test-host-write-codex test-codex-dispatch-continuation test-host-write-parity test-core-schemas test-layer-boundaries test-pm-scripts test-run-tests test-state-store test-state-layout-parity)
+suites=(lint-agents lint-scripts lint-script-domain-inventory lint-portable-repo-paths test-lint-shellcheck test-script-domain-inventory test-lint-portable-repo-paths test-lint-frontmatter test-commands test-check-docs-freshness test-guards test-migrate test-install test-doctor test-pmctl-dispatch test-pmctl-context test-pmctl-memory test-pr-gate test-pr-gate-profile test-pmctl-operation test-host-manifest test-host-write-codex test-codex-dispatch-continuation test-host-write-parity test-core-schemas test-layer-boundaries test-pm-scripts test-run-tests test-state-store test-state-layout-parity)
 for arg in "$@"; do
   if [[ "$arg" == --list ]]; then printf '%s\n' "${suites[@]}"; exit 0; fi
 done
@@ -228,6 +231,23 @@ case_evidence_contract_maps_to_runner_regression() {
   fi
 }
 
+case_gate_assurance_policy_maps_gate_consumers() {
+  local name=gate-assurance-policy-maps-gate-consumers repo out status=0 args
+  args="$TMP_ROOT/$name.args"
+  repo="$(make_fixture "$name")"
+  out=$(RUN_TESTS_ARGS_LOG="$args" "$repo/tests/bin/run-tests.sh" \
+    --path core/policy/gate-tiers.tsv --path core/policy/gate-modes.tsv \
+    --path core/policy/gate-pass-kinds.tsv --list 2>&1) || status=$?
+  if [[ "$status" -eq 0 && "$out" == *"test-pr-gate"* &&
+        "$out" == *"test-pr-gate-profile"* && "$out" == *"test-core-schemas"* &&
+        "$out" == *"test-layer-boundaries"* && "$out" != *"coverage gaps"* &&
+        ! -e "$args" ]]; then
+    pass "$name"
+  else
+    fail "$name" "status=$status out=$out"
+  fi
+}
+
 case_high_fanout_escalates_full() {
   local name=high-fanout-escalates-full repo out status=0 args
   args="$TMP_ROOT/$name.args"
@@ -402,6 +422,7 @@ case_skill_mapping_uses_registered_frontmatter_suite
 case_guard_family_maps_to_guard_suite
 case_prompt_context_timeout_contract_maps_all_consumers
 case_evidence_contract_maps_to_runner_regression
+case_gate_assurance_policy_maps_gate_consumers
 case_high_fanout_escalates_full
 case_repeated_high_fanout_escalation_succeeds
 case_unknown_path_fails_without_test_evidence