docs: prioritize steward journeys and small-team golden-query evaluation

[huangruiteng]

Goal And Delivered Outcome

Steward planning prioritized answer presentation before proving that an ordinary request reaches a responsible worker and returns a usable result. This proposal makes creation/connection, responsible dispatch and real small-team coordination explicit P0 outcomes, with portfolio management and distributed scale following at P1/P2.

A public-safe pack defines 17 natural query families, the required initial context, independent acceptance observations, recovery/correction variants and paired outcome/attention/cost measures. The overall roadmap and manager/presentation RFCs now point to the same evaluation targets in both language versions. It reuses R1–R3, G1–G4 and M1–M3/A1–A24; intended base: main.

Broad owner discovery is a P0 default: the registered local inventory and already-authorized connected sources remain searchable beyond the first page. Discovery, evidence access, delegation grants and execution readiness are separate; a narrow recipient allowlist must not imply that no matching Agent exists. Shared audiences retain their own visibility boundary. These are target defaults, not a claim of shipped runtime behavior.

Scope And Continuation

Planning/evaluation scope is complete. Runtime implementation and live golden-query acceptance remain open: no passing product score, new protocol, capability, runner or model call is claimed. Implementation continues through existing creation/Chat, directory/binding, collaboration/artifact and return owners. Small-team acceptance includes parallel joins, peer help, disagreement, independent review and receiver adoption across two real cycles.

Future-facing pass: replace the report-first dependency and stale #4978 proposed status; keep one case/measurement pack rather than duplicate acceptance rules across RFCs. Canonical task planning is maintained outside the public repository. Private conversations and operating state are excluded.

Validation

• Tested revision: 5c200fa
• Run state: finished
• Input classes: public_fixture, authorized_private_read_only

Check kind	Result	Public-safe evidence / limitation
static	passed	uv run --extra test python examples/docs-governance-smoke.py; documentation placement and links
static	passed	uv run --extra test python examples/docs-asset-integrity-smoke.py; 7 existing assets verified
static	passed	git diff --check and explicit changed-path public-boundary scan; zero public errors. Two unrelated existing Goal-state warnings were not altered
manual	passed	Reviewed bilingual priority/ownership parity, natural query wording, positive/negative outcomes, public/private separation and existing milestone reuse
static	passed	Exact-scope change-quality receipt cqr_1505ef9092f41e6de3c0, 9 files, fingerprint 1505ef9092f41e6de3c0fef02d3c4de1784b72033521dd00ed671427e6bc3161; safe fix permitted/not applied, zero repair passes, zero blockers/warnings/advisories
integration	prior revision only	On prior head 943ab8b, risk-selected premerge: 19 selected checks passed. The latest docs-only refinement reran the focused documentation/boundary checks above; the full selection was not repeated. Initial vocabulary check could not load missing worktree Node dependencies; installed with npm ci --ignore-scripts, then reran the full selection successfully. No source changes were needed
real_entrypoint	not_run	All live query cases remain unqualified. Private read-only observations informed generalized cases; they do not certify current runtime causes or successful repairs

Coverage: docs-only diff; runtime/frontend/backend execution is unchanged. The proposed 30% reduction in avoidable coordination is a future evaluation target, not a measured improvement. Later cross-host/scale cases require their own frozen fixture packs and authorized resources.

Frontend / Visual Evidence

• UI impact: none. Markdown evaluation and RFC content only; no product chrome, primary navigation, rendering code or packaged assets changed.
• Source data: public_fixture. No private screenshots or message bodies are attached.

Type of Change

• Documentation update

LoopX Area

• Public docs or presentation surface (README, protocols, dashboard)

Technical Direction

Operator surface and IM integration; R1–R3/G1 supported path first, R3/R4 portfolio management next, R6/G3 then R7/G4 separately qualified.

Shared-authority RFC fixture impact

N/A: no provider, transaction, compatibility or promotion change.

Boundary Checklist

• No private state, credentials, raw traces, internal links or local machine paths in the diff or PR.
• No duplicated maintainer-owned benchmark work or live model experiment.
• Scope limited to the product evaluation plan and owning RFCs.
• UI impact is none.
• Every commit includes a DCO sign-off.

[huangruiteng]

Approval conclusion (author-owned PR; GitHub blocks formal self-approval)

动机

用户需要把“创建、找负责人、真实小团队协作、结果回到原入口”排到同一产品目标里。旧文档缺少一套简短请求和可独立检查的结果标准，容易让目录、收件箱回执或漂亮的回答被误当成完整交付。本 PR 将这些目标明确为 P0，同时保留后续能力的层次。

改动思路

以一份公开安全的 golden-query 集作为评测入口，再让整体 roadmap、管家 RFC、共用对话界面 RFC 双语引用它。它复用既有 R/G/M/A 里程碑和 Todo owner，不创建新协议或状态机。正向路径是冻结场景、运行基线与候选、检查负责人实际执行和原对话回传；失败路径明确保留未验收状态。

具体改动

• docs/product/use-cases/steward/golden-queries.md 定义 17 类自然请求、P0/P1/P2 分层、两轮协作和独立质量/注意力度量；创建、负责人发现、采用、纠偏、停止、恢复都有反例。
• 中英 roadmap 把入口与负责人路由、2–3 人真实协作列为 P0；中英管家 RFC 补足宽范围发现、委派准入及真实回传；展示 RFC 强调成果、普通进度、必要决策和首次请求保留。Steward use-case 索引提供直接入口。

关键内容讲解

GQ03 把“找 PR review 负责人”验收到因果检查、实际处理和回传，明确基线 CI 失败不是自动批准。GQ05/GQ11–GQ13 要求接收方使用修订版本、独立复核、处理分歧并完成两轮；一条 inbox ACK 不算协作。宽范围发现与证据读取、委派、执行就绪分别判断，外部受众不能继承主人的可见范围。

对主干的风险

主要风险是把规划误读为现有能力。文件明确标明所有真实案例尚未验收、30% 注意力下降只是预设目标，跨主机和规模场景需要另行冻结资源。完整 diff 无运行时代码和状态迁移；本地文档治理、资源完整性、空白和签署检查通过。未跑真实 Agent，因此不能据此作产品发布结论。

我的整体评价

对长期协作和用户体验的规划都有正向增量：它让后续 PR 必须证明实际执行、采用和回传，也避免为了减少消息数牺牲结果质量。没有阻塞发现。文档切片完整且可独立撤回；残余风险是评测尚未实施，需由既有 R2/R3/A24 owner 用真实入口继续验收。

English verdict: APPROVE - exact head 5c200fa30835fd6f7699a82e848fea58cb5161bb; docs-only P0 evaluation contract, focused documentation checks passed, live golden-query acceptance remains open.

[huangruiteng]

Approval conclusion (author-owned PR; GitHub blocks formal self-approval)

动机

用户需要把“创建、找负责人、真实小团队协作、结果回到原入口”排到同一产品目标里。旧文档缺少一套简短请求和可独立检查的结果标准，容易让目录、收件箱回执或漂亮的回答被误当成完整交付。本 PR 将这些目标明确为 P0，同时保留后续能力的层次。

改动思路

以一份公开安全的 golden-query 集作为评测入口，再让整体 roadmap、管家 RFC、共用对话界面 RFC 双语引用它。它复用既有 R/G/M/A 里程碑和 Todo owner，不创建新协议或状态机。正向路径是冻结场景、运行基线与候选、检查负责人实际执行和原对话回传；失败路径明确保留未验收状态。

具体改动

• docs/product/use-cases/steward/golden-queries.md 定义 17 类自然请求、P0/P1/P2 分层、两轮协作和独立质量/注意力度量；创建、负责人发现、采用、纠偏、停止、恢复都有反例。
• 中英 roadmap 把入口与负责人路由、2–3 人真实协作列为 P0；中英管家 RFC 补足宽范围发现、委派准入及真实回传；展示 RFC 强调成果、普通进度、必要决策和首次请求保留。Steward use-case 索引提供直接入口。

关键内容讲解

GQ03 把“找 PR review 负责人”验收到因果检查、实际处理和回传，明确基线 CI 失败不是自动批准。GQ05/GQ11–GQ13 要求接收方使用修订版本、独立复核、处理分歧并完成两轮；一条 inbox ACK 不算协作。宽范围发现与证据读取、委派、执行就绪分别判断，外部受众不能继承主人的可见范围。

对主干的风险

主要风险是把规划误读为现有能力。文件明确标明所有真实案例尚未验收、30% 注意力下降只是预设目标，跨主机和规模场景需要另行冻结资源。完整 diff 无运行时代码和状态迁移；本地文档治理、资源完整性、空白和签署检查通过；新基线的 19 项风险选择合并前检查全部通过，质量凭据 cqr_dc60b1f7885721fc7044 有效，无失败或跳过。未跑真实 Agent，因此不能据此作产品发布结论。

我的整体评价

对长期协作和用户体验的规划都有正向增量：它让后续 PR 必须证明实际执行、采用和回传，也避免为了减少消息数牺牲结果质量。没有阻塞发现。文档切片完整且可独立撤回；残余风险是评测尚未实施，需由既有 R2/R3/A24 owner 用真实入口继续验收。

English verdict: APPROVE - exact head 2b27e1618fd68fb5c4f1809e023f141482b200b3; docs-only P0 evaluation contract, focused documentation checks passed, live golden-query acceptance remains open.