perf: load conditional-path libs on demand instead of inlining every branch

PRD.md

@@ -50,7 +50,7 @@ Aligned with [GOALS.md](./GOALS.md)'s Core Goals:
 | FR-2 | The system MUST support installing into an explicit subset of environments via `--env <envs>` (comma-separated, case-insensitive, trimmed), including the aliases `gemini`/`agy` → `antigravity`. | Must | `--env CLAUDE,agy` installs into the claude and antigravity targets only. |
 | FR-3 | The system MUST support installing/uninstalling a filtered subset of commands by name, accepting both bare (`push`) and `do:`-prefixed (`do:push`) forms. | Must | `npx slash-do@latest push pr` installs only the push and pr commands. |
 | FR-4 | The system MUST transform each source command file into the native format of its target environment — subdirectory layout for Claude Code, flat `do-<name>.md` for OpenCode, directory-per-skill (`SKILL.md`) for Antigravity/Codex/Grok — without per-environment hand-authoring. | Must | A single source file in `commands/do/` produces a correctly formatted, working command/skill file in every installed environment. |
-| FR-5 | For environments without `!cat` file-inclusion support, the system MUST inline referenced `lib/*.md` content directly into the transformed command file (recursively, with cycle termination) rather than leaving a broken path reference. | Must | No dangling `~/.claude/lib/<name>.md` references appear in Agent Skills output, even when the referenced file is missing. |
+| FR-5 | For environments without `!cat` file-inclusion support, the system MUST make every referenced `lib/*.md` reachable — inlining content into the transformed command file (recursively, with cycle termination), or, for a mutually-exclusive lib the run selects at most one of, writing it into the skill's own `lib/` directory and citing it by a relative path the agent reads on demand. Neither may leave a broken path reference. | Must | No dangling `~/.claude/lib/<name>.md` references appear in Agent Skills output, even when the referenced file is missing; every `lib/<name>.md` cited in a SKILL.md resolves to a file installed beside it. |
 | FR-6 | Re-running install MUST be idempotent — unchanged files report up to date, changed files are updated in place, and hooks/config are not duplicated. | Must | Two consecutive installs with no source changes report 0 updates on the second run. |
 | FR-7 | `--dry-run` MUST preview changes without writing, creating, or deleting any file, directory, hook registration, or config entry. | Must | A dry-run install on a clean host leaves the filesystem byte-for-byte unchanged. |
 | FR-8 | `--list` MUST show all commands and their install status per environment without making changes. | Must | Output lists every command with an installed/not-installed/outdated status per detected environment. |

README.md

@@ -454,7 +454,7 @@ npx slash-do@latest push pr release           # install specific commands only
   +------------------+
   |   Transformer    |  Converts format per environment:
   |                  |  - YAML frontmatter (Claude, OpenCode)
-  +------------------+  - Agent Skills / SKILL.md with inlined libs (Antigravity, Codex, Grok Build)
+  +------------------+  - Agent Skills / SKILL.md + bundled lib/ (Antigravity, Codex, Grok Build)
        |
        v
   +------------------+

install.sh

@@ -176,7 +176,7 @@ LIBS=(
   empty-array-expansion enhance-loop epic-children
   finding-disposition fix-regression-guard
   gh-host github-reviewer-loop graphql-escaping
-  local-agent-review-loop model-tiers multi-reviewer-loop ollama-review-loop
+  local-agent-review-loop model-tiers multi-reviewer-loop next-swarm ollama-review-loop
   per-finding-root-cause plan-id-format plan-issue-mode
   post-review-doc-recommendations remediation-agent-template
   review-agent-selection review-config-defaults review-convergence-gate

lib/multi-reviewer-loop.md

@@ -59,7 +59,7 @@ This is the default path. Iterate `REVIEW_AGENTS` in order, running each reviewe
 
 1. **Print a banner**: `--- Review pass {n}/{N}: {REVIEW_AGENT} ---`
 2. **Capture baseline**: `PASS_START_SHA=$(git rev-parse HEAD)` so the wrapper can tell whether this reviewer changed anything (independent of the inner loop's own tracking).
-3. **Dispatch** to the matching single-reviewer loop. The loop file lives under the host CLI's lib directory (`~/.claude/lib/` for Claude, `~/.config/opencode/lib/` for OpenCode — slashdo's installer rewrites command-spec `!cat` references for each env, so use the same lib basename in whichever env this wrapper executes. For Antigravity and Codex there is no separate lib path: slashdo inlines the loop bodies directly into the installed skill, so the dispatch targets below are already present in-context rather than at a file path):
+3. **Dispatch** to the matching single-reviewer loop. The loop file lives under the host CLI's lib directory (`~/.claude/lib/` for Claude, `~/.config/opencode/lib/` for OpenCode) or bundled alongside the skill under `lib/` (for Antigravity, Codex, Grok — read on demand per the directive in this skill):
    Forward this entry's resolved `{MAX_ITERATIONS}` and `{MAX_EXPLICIT}` (from the cap-resolution step) to **every** target below — that pair is what carries a `~max=<n>` into the reviewer; forward `{ENTRY_EFFORT}` as `{REVIEW_EFFORT}` / `{OLLAMA_EFFORT}` (empty if unset):
    - `copilot` → `{LIB_DIR}/copilot-review-loop.md` (forward `{GH_HOST}`, and this entry's `{MAX_ITERATIONS}` as its `{REVIEW_ITERATIONS}` iteration cap)
    - `@<login>` → `{LIB_DIR}/github-reviewer-loop.md` (forward `{GH_HOST}`, this entry's `{REVIEWER_LOGIN}`, and this entry's `{MAX_ITERATIONS}` as its `{REVIEW_ITERATIONS}` iteration cap)