2026-10-18 18:00:19 +00:00
3 changed files with 278 additions and 3 deletions
@@ -0,0 +1,216 @@
+# feature/norm-remediation-u7 任务卡
+
+任务主记录：PR #622 https://git-home.u1680832.nyat.app:30319/weichen/Agent-System-Dev/pulls/622；状态：已完成（合入前闭合）；owner：weichen
+
+- 权威仓：ssh://git@git-ssh.u1680832.nyat.app:31411/weichen/Agent-System-Dev.git；目标分支：main；R0（initial_base_sha）= eef7e5e778cabcb6216f74b74213a3d7511f383b
+- 工作区：/home/weichen/projects/Agent-System-Dev-wt-feature-norm-remediation-u7（worktree）；分支：feature/norm-remediation-u7
+- 分支类型：child；initiative_id：feature-norm-remediation；root_target：main；immediate_target：feature/norm-remediation；accepting_requirements：false；parent_branch：feature/norm-remediation；parent_head_sha：c6fe5adba34d7266e7bcfbb9a2c64a0ae61e9b8e；parent_pr：无；parent_source：origin
+- 需求分流：免（父 initiative 子任务登记已列出 feature/norm-remediation-u7；parent_head_sha=c6fe5adba34d7266e7bcfbb9a2c64a0ae61e9b8e）
+
+- 授权与交付终点：用户追加并确认 U7：完善 subagent/CLI 的有界任务、容量规划、模型分工、阶段交接与测试要求；本子任务经 Review/CI 合入 initiative。范围限 `AGENT-EXECUTION.md`、`policy/delivery.md`；通用规则已由 Guidelines PR #11 合入，最终 pin/采纳归 U6。本任务不部署、不改平台或执行器架构。
+- 交付模式：goal-first（目标、范围和交付终点一次确认；范围内连续推进）
+- 实施执行：agent-per-unit（配置：AGENT-EXECUTION.md）
+- 合入执行：agent（已授权交付范围内，Review 通过、冲突解决且当前 HEAD 的 CI 通过后合入）
+- 流程依据：仓库根目录 policy/delivery.md（归档后仍按仓库根目录定位）
+
+## 目标与非目标
+
+目标：工作包适合较小/未知容量模型，以一个可独立验证的结果、最小背景和明确验收派发；阶段完成自动持久交接，下一 CLI 会话全新，原生压缩仅在支持时可选；具体规划/执行模型偏好归项目本地配置，requested/effective/不可观测状态如实区分。未知容量使用 100K 显式规划假设，给实现/迭代/测试保留空间；不新增控制器、Schema 或失败门禁。
+
+共享基线：Guidelines PR #11，merge `a15db5085585f199c3c399824f6cf24c3144bbd5`（draft.14）；独立 Review、schema/config 校验与 CI run 3978 均通过。正文/模板和 Review 修复 transport 分别见 `/tmp/guidelines-u7-receipt.json`、`/tmp/guidelines-u7-review-repair-receipt.json`；worker 私有索引 intent-to-add 曾被撤销，为执行偏差，不证明提示具备工具强制力。
+
+## Agent 实施收据
+
+路由与 fail-stop 规则见根目录 AGENT-EXECUTION.md 与 policy/delivery.md；每次尝试一行，原始日志留在源码外收据路径。
+
+| 单元 ID | 预期范围 | 配置摘要 | 角色 | 候选顺序 | attempt | CLI/模型 | 终态 | transport | fallback | 源码外收据路径 |
+|---|---|---|---|---|---|---|---|---|---|---|
+| norm-u7-dev-bounded-cli-routing | AGENT-EXECUTION.md + policy/delivery.md | caefcb72 (prior config) | implementation | cursor, agy, pi | 1 | cursor-agent-stable / auto | success | e16a75c5111c | 0 | /tmp/norm-u7-dev-receipt.json |
+| norm-u7-dev-independent-review | independent fixed-diff/config review | run receipt | review | pi, agy, cursor, codex | 1 | pi / minimax-cn/MiniMax-M3 (actual event) | success | 892cf59c5075 | 0 | /tmp/norm-u7-dev-review-receipt.json |
+
+## 同步记录
+
+| 节点 | sync_base_sha | tested_head_sha | checked_at |
+|---|---|---|---|
+| 启动 | eef7e5e778cabcb6216f74b74213a3d7511f383b | c6fe5adba34d7266e7bcfbb9a2c64a0ae61e9b8e | 2026-10-18T17:35:52Z |
+
+## 证据（Evidence）
+
+<!-- CMD="<命令>" make evidence 写入；手抄/转写输出视为无效证据（AGENTS.md 证据与验证门禁）。 -->
+| 时间(UTC) | 命令 | 被测HEAD | exit | 来源 | sha256(16) |
+|---|---|---|---|---|---|
+
+## 结果与验证
+
+- 结果：共享有界工作包与交接标准已合入 Guidelines PR #11（draft.14）；本仓补齐工作包最小信息、100K 保守假设、阶段交接和 requested/effective 区分。PI 请求已明确为 `minimax-cn/MiniMax-M3`，独立 Review 实际事件也报告同 provider/model；Cursor auto，agy Flash 偏好保持 NOT_VERIFIED。普通上下文压力不触发扩围停止；无自动压缩或未知容量不阻断其他授权工作。
+- 验证：host 在 `e16a75c5111c861afea11458194ac308f81ca99a` 运行配置校验、6 项 agent-exec 入口测试与 diff-check 均通过（SELF）；独立 fresh PI Review 重放固定 diff/配置，无阻塞 finding。两个 CLI run 均为新进程和独立隔离工作区，未使用 resume/continue；session ID 未单独提取时记 not-observable，独立性依据 run/attempt 及新进程，不伪造 ID。
+- 影响与剩余项：U6 才推进 Guidelines pin/采纳记录；当前旧 pin 不宣称已经采用 draft.14。未知/已知容量例子是规范一致性检查，不证明实际模型最大容量或质量提升；同任务真实模型前后质量基准 NOT_RUN（本次交付是规范与配置，未建设评测平台，也未宣称质量提升的实测数值）。端点/登录/压缩与并发强制能力均不作无证据断言。
+
+## 合入与清理
+
+- Review 记录（stage=review，fresh_session=true，model=minimax-cn/MiniMax-M3 as configured for PI，session ID=not-observable）：独立 Review session；base=c16665365bbbf3b733b034628f102db90d985637；head=e16a75c5111c861afea11458194ac308f81ca99a；范围限 diff AGENT-EXECUTION.md + policy/delivery.md。
+  - 命令与对象：`git diff c16665365bbbf3b733b034628f102db90d985637..e16a75c5111c861afea11458194ac308f81ca99a -- AGENT-EXECUTION.md policy/delivery.md` → 98 行差异（AGENT-EXECUTION.md +41/-3，policy/delivery.md +24/-0）；`scripts/agent-exec validate --repo .`（deps/development-manager 未初始化，回退到精确钉住 dm 二进制 `$HOME/.cache/agent-system-dev/agent-exec/dm-server-67784f43eb0430bf596c2cda685cbb4946d2e032` 与 index 钉住 `67784f43eb0430bf596c2cda685cbb4946d2e032` 完全一致；未写规范仓；binary version=0.1.0-dev）→ `validate: ok`，exit=0。
+  - 结论：无 blocker。机器块（`execution.fallback_to_host=never`、`max_attempts_per_profile=3`、`emergency_profile=codex`、`roles.implementation/repair/test/review` 候选顺序、超时/输出字节）逐项与 base 一致；PI `model` 由 `auto` 改为显式 `minimax-cn/MiniMax-M3`，Cursor 保持 `auto`，agy 保持 `auto` 并在 prose 明示 Gemini 3.8 Flash 为 `NOT_VERIFIED`（未观测模型 ID，未声称生效）；新增 prose 段只描述本仓做法并显式声明「本节不新增门禁，不放宽凭据、范围、未知副作用与 CI 边界」与「Role lists, attempt limits, timeouts, output limits, the single Codex emergency attempt, stage isolation and fallback_to_host: never are unchanged」。
+  - 观测性：`requested` 模型是提示/配置意图，`effective` 以运行事件信封为准，CLI 不报告时记 `not-observable`；未知容量、agy 缺登录、缺自动压缩均登记事实，不作为全局 fail-close（仅真实扩围走既有 `scope-expansion` fail-stop）。现有 scope/credential/unknown-side-effect fail-stops 未删未弱。
+  - 单元包/容量：bounded packet JSON 的 `goal/constraints/verification` 已含一个结果与验收、允许路径、最小必读、前置 SHA、直接验证命令、未知容量按保守 100K 假设并为执行留余量；100K/200K/300K/500K 仅为规划参考，未声明任何 provider 能力；不新增 schema/字段/控制器。
+  - 阶段交接：里程碑持久交接写入既有任务记录，下一 CLI 阶段或单元以全新会话启动，只读交接与最小必读，不为交接等人工。
+  - 上下文压力：仅在 provider 支持且阶段完成后做上下文压缩；不支持不阻塞；普通压力走交接 + host 重新打包。
+  - 非阻塞观察（不阻塞交付）：`AGENT-EXECUTION.md`/`policy/delivery.md` 中新增的相对链接指向 `deps/development-guidelines/standards/agent-execution.md` 与 `deps/development-guidelines/workflows/handoff.md`；当前 `deps/development-guidelines` gitlink 仍为 `2688fd2be6a52c0bf110e7e6366b6f291ab5e4eb`（未在 U7 范围 bump），文件存在与否须 U6 pin 时再核，prose 未声称当前 pin 已采纳 Guidelines PR #11（merge `a15db5085585f199c3c399824f6cf24c3144bbd5`），仅引导后续采纳。`policy/delivery.md` 锚 `#有界单元包上下文与阶段交接` 与新章节 `### 有界单元包、上下文与阶段交接` 在 GFM 风格锚归一化（移除中文标点）下等价。`agy` 段仅将「Gemini 3.8 Flash」记为请求，未声明能力，符合「不发明模型能力」要求。
+  - 不主张 CI 通过；未运行任何 live API/模型探测；未改 source/config/index/refs；未开 PR/部署/派遣同伴。审阅者身份：本独立 session（PI / minimax-cn/MiniMax-M3，session ID not-observable，stage=review）。重放依据：`git rev-parse HEAD` 与 `git ls-files -s deps/development-manager` 即可复现。
+- 冲突记录：已同步包含 U4 的 initiative `825e4fc`，无冲突；U7 代码/配置相对已审实现 `e16a75c` 未再改变。host 在合成 head `3da20d4` 校验配置并重放入口+事实测试 14 项，均通过（SELF，receipt 已导入），未重跑与此无关的完整套件。
+- CI 记录：NOT_RUN（当前 HEAD 完整 SHA、全部必需检查、运行链接、执行身份及 GATE 结果）
+- 合入记录：NOT_RUN（交付授权范围、PR、被核验 HEAD、合入 SHA、目标分支及内容核验）
+- 清理记录：NOT_RUN（授权依据、写者退出确认、worktree/分支/任务卡处置）
+
+<!-- Local receipt: SELF. Digests detect corruption; they do not authenticate a reviewer or CI. -->
+```json
+{
+  "summary_schema": 1,
+  "source": "SELF",
+  "timestamp": "2026-10-18T17:34:44.780514+00:00",
+  "command": "make precheck",
+  "exit_code": 0,
+  "cwd": "/home/weichen/projects/Agent-System-Dev",
+  "before_head": "eef7e5e778cabcb6216f74b74213a3d7511f383b",
+  "after_head": "eef7e5e778cabcb6216f74b74213a3d7511f383b",
+  "branch": "main",
+  "target_branch": "main",
+  "target_head": "eef7e5e778cabcb6216f74b74213a3d7511f383b",
+  "root_target": "main",
+  "immediate_target": "main",
+  "before_tracked_diff_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "after_tracked_diff_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "output_tail": [
+    "make[1]: Entering directory '/home/weichen/projects/Agent-System-Dev'",
+    "[repo]   branch=main head=eef7e5e778",
+    "[remote] origin/main=eef7e5e778 ahead=0 behind=0",
+    "[tree]   clean submodules_drift=0 uninit=12",
+    "[warn] 12 个子模块未初始化——构建/测试前需物化（不阻断写入）",
+    "[worktrees] linked=7",
+    "[recycle] merged un-recycled branches: feature/layered-python-env",
+    "[initiatives] local cards (not authoritative): none",
+    "[hooks]   core.hooksPath=scripts/hooks（pre-commit/post-merge 实际接线）",
+    "[facts]   policy/facts.md 6 条事实均在有效期内",
+    "[required-reading] core/baseline.md, workflows/delivery.md §1, policy/delivery.md",
+    "[R0] eef7e5e778（完整：eef7e5e778cabcb6216f74b74213a3d7511f383b）",
+    "[worktree] before-sync: recursive component checks passed",
+    "[verdict] READY — 保留本次 precheck receipt 后建任务工作区：PRECHECK_RECEIPT=/tmp/task-precheck.json make task-start NAME=feature/your-branch",
+    "make[1]: Leaving directory '/home/weichen/projects/Agent-System-Dev'"
+  ],
+  "output_sha256": "8ebd14a4e8639e7dffb04336ad69df1bb8f80b36d3f9cf1c0875966b3503b3f4",
+  "receipt_sha256": "595376851618682c9df8e7c637c465e9950237249d07bf36ea0726026961fe06",
+  "receipt_path": "/home/weichen/.cache/agent-system-dev/evidence/595376851618682c9df8e7c637c465e9950237249d07bf36ea0726026961fe06.json"
+}
+```
+
+<!-- Local receipt: SELF. Digests detect corruption; they do not authenticate a reviewer or CI. -->
+```json
+{
+  "summary_schema": 1,
+  "source": "SELF",
+  "timestamp": "2026-10-18T17:39:09.639625+00:00",
+  "command": "scripts/agent-exec validate --repo . && python3 -m unittest tests.test_agent_exec_entry && git diff --check",
+  "exit_code": 0,
+  "cwd": "/home/weichen/projects/Agent-System-Dev-wt-feature-norm-remediation-u7",
+  "before_head": "e16a75c5111c861afea11458194ac308f81ca99a",
+  "after_head": "e16a75c5111c861afea11458194ac308f81ca99a",
+  "branch": "feature/norm-remediation-u7",
+  "target_branch": "main",
+  "target_head": "eef7e5e778cabcb6216f74b74213a3d7511f383b",
+  "root_target": "main",
+  "immediate_target": "feature/norm-remediation",
+  "before_tracked_diff_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "after_tracked_diff_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "output_tail": [
+    "validate: ok",
+    "......",
+    "----------------------------------------------------------------------",
+    "Ran 6 tests in 0.389s",
+    "",
+    "OK"
+  ],
+  "output_sha256": "8fde6c3d2592daa001b2e9856870a9725c40bcd8c52e7e56c58852005c56eab7",
+  "receipt_sha256": "1478aa40c6d51ff8198bf6517ba5fa4a7fd51f33a7112d52057e771f346cf248",
+  "receipt_path": "/home/weichen/.cache/agent-system-dev/evidence/1478aa40c6d51ff8198bf6517ba5fa4a7fd51f33a7112d52057e771f346cf248.json"
+}
+```
+| 同步（记录前对象） | 825e4fc98f30da4950a48d15d59f5338206726bf | 66696bfc436a4adadf70fe7ecfc6ab4cda631d65 | 2026-10-18T17:48:16Z |
+
+<!-- Local receipt: SELF. Digests detect corruption; they do not authenticate a reviewer or CI. -->
+```json
+{
+  "summary_schema": 1,
+  "source": "SELF",
+  "timestamp": "2026-10-18T17:47:48.776787+00:00",
+  "command": "make task-sync",
+  "exit_code": 0,
+  "cwd": "/home/weichen/projects/Agent-System-Dev-wt-feature-norm-remediation-u7",
+  "before_head": "2de3d398ccc1db9d235238016a12755b4ea989f2",
+  "after_head": "3da20d47b5a4a0732effad4992daf17acd2af944",
+  "branch": "feature/norm-remediation-u7",
+  "target_branch": "main",
+  "target_head": "eef7e5e778cabcb6216f74b74213a3d7511f383b",
+  "root_target": "main",
+  "immediate_target": "feature/norm-remediation",
+  "before_tracked_diff_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "after_tracked_diff_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "output_tail": [
+    " policy/facts.md                              |   9 +-",
+    " scripts/facts-verify.sh                      |  79 ++++++",
+    " tests/test_facts_verify.py                   | 399 +++++++++++++++++++++++++++",
+    " 4 files changed, 746 insertions(+), 2 deletions(-)",
+    " create mode 100644 .task/archive/feature-norm-remediation-u4.md",
+    " create mode 100644 tests/test_facts_verify.py",
+    "[worktree] before-update: recursive component checks passed",
+    "[worktree] materialize: recursive component checks passed",
+    "[worktree] after-sync: recursive component checks passed",
+    "admit: ok",
+    "[evidence] WARN JSON receipt line 66: unverified: recorded component state was not at its pinned SHA ——关键检查须重跑并以 make evidence 重新写入",
+    "[evidence] WARN JSON receipt line 107: recorded HEAD predates non-card changes ——关键检查须重跑并以 make evidence 重新写入",
+    "[synced] sync_base=825e4fc98f30da4950a48d15d59f5338206726bf tested_head=3da20d47b5a4a0732effad4992daf17acd2af944 checked_at=2026-10-18T17:48:16Z changed=1 sync_target=feature/norm-remediation",
+    "[next] 跑正式验收（make test 等），证据写入任务卡",
+    "make[1]: Leaving directory '/home/weichen/projects/Agent-System-Dev-wt-feature-norm-remediation-u7'"
+  ],
+  "output_sha256": "0397ae866aa211d57b600b60355563e5dc4ebb1d09569ef0cc607dfbe5bd1f27",
+  "receipt_sha256": "bcb38dab3f750867a93de90576aa0d39f7fd0501fb0557c085a6d26cf1a29779",
+  "receipt_path": "/home/weichen/.cache/agent-system-dev/evidence/bcb38dab3f750867a93de90576aa0d39f7fd0501fb0557c085a6d26cf1a29779.json"
+}
+```
+
+<!-- Local receipt: SELF. Digests detect corruption; they do not authenticate a reviewer or CI. -->
+```json
+{
+  "summary_schema": 1,
+  "source": "SELF",
+  "timestamp": "2026-10-18T17:49:37.631093+00:00",
+  "command": "scripts/agent-exec validate --repo . && python3 -m unittest tests.test_agent_exec_entry tests.test_facts_verify && git diff --check",
+  "exit_code": 0,
+  "cwd": "/home/weichen/projects/Agent-System-Dev-wt-feature-norm-remediation-u7",
+  "before_head": "3da20d47b5a4a0732effad4992daf17acd2af944",
+  "after_head": "3da20d47b5a4a0732effad4992daf17acd2af944",
+  "branch": "feature/norm-remediation-u7",
+  "target_branch": "main",
+  "target_head": "eef7e5e778cabcb6216f74b74213a3d7511f383b",
+  "root_target": "main",
+  "immediate_target": "feature/norm-remediation",
+  "before_tracked_diff_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "after_tracked_diff_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+  "output_tail": [
+    "validate: ok",
+    "..............",
+    "----------------------------------------------------------------------",
+    "Ran 14 tests in 4.466s",
+    "",
+    "OK"
+  ],
+  "output_sha256": "0abf86ce43ea39bd4b51a9be1a3ca8bc47531f3e75e13af28f7f1f4d6452bc19",
+  "receipt_sha256": "3fb74abf04253157034696850c14561fdee89c6a47e3da22aaea6a5ef419ef70",
+  "receipt_path": "/home/weichen/.cache/agent-system-dev/evidence/3fb74abf04253157034696850c14561fdee89c6a47e3da22aaea6a5ef419ef70.json"
+}
+```
+
+## 闭合记录（合入前折叠）
+
+- 本卡在 PR #622（https://git-home.u1680832.nyat.app:30319/weichen/Agent-System-Dev/pulls/622）合入前闭合并归档，随该 PR 一同合入。
+- 该 PR 的合并提交即本任务的落地记录。
+- 合入后无需回填，也无需单独的闭合 PR。
@@ -18,8 +18,8 @@ increment host fallback authorizations. Safe exhaustion without an emergency
 candidate ends `unavailable` or `blocked`, and that result does not authorize
 the host to write the unit. The legacy value `all-unavailable` remains
 parseable and likewise grants no host write authority. Cursor uses
-`cursor-agent-stable`. `model: auto` lets each CLI use its configured default
-model; PI therefore uses the operator-selected MiniMax default. The
+`cursor-agent-stable`. `model: auto` lets a CLI use its configured default
+model; any other value is passed to that CLI as its model selector. The
 `integration` block declares the host-exempt paths and the transport, admission,
 and cleanup entry points.

@@ -60,7 +60,7 @@ profiles:
  pi:
    adapter: pi
    executable: pi
-    model: auto
+    model: minimax-cn/MiniMax-M3
    enabled: true

  codex:
@@ -78,3 +78,38 @@ integration:
  cleanup:
    command: make task-finish
 ```
+
+## Model preferences (project choice)
+
+Generic rules for bounded packets, context budgeting, phase handoff and
+requested-versus-effective model identity live in the shared
+[standards/agent-execution.md](deps/development-guidelines/standards/agent-execution.md)
+and [workflows/handoff.md](deps/development-guidelines/workflows/handoff.md);
+project procedure is in
+[policy/delivery.md](policy/delivery.md#有界单元包上下文与阶段交接). This
+section records only this project's choices.
+
+- Planning, design and implementation plans: Codex or Claude when available,
+  in the host or a fresh CLI session. This is a preference, not a routing role.
+- Bounded implementation, repair and test units: the existing normal
+  candidates in the configured role order above. Requested models:
+  - PI: `minimax-cn/MiniMax-M3`, passed as `--model provider/id`. PI
+    actual events have reported `minimax-cn` / `MiniMax-M3` and the local
+    offline catalog lists that profile; each run still records the effective
+    model from that run's events.
+  - Cursor: `auto` (Cursor Auto).
+  - agy: Gemini 3.8 Flash is requested, but its exact selectable model ID and
+    effective identity are `NOT_VERIFIED` (the `agy` models query needs a
+    login that was not available). The profile therefore stays `model: auto`
+    until an observed model ID exists; do not claim Gemini 3.8 Flash is active.
+- Assigning a different candidate or model to a unit needs a concrete,
+  task-specific reason recorded in the task record, and stays within the
+  existing roles and candidate lists. It is not a silent permission or routing
+  change. Role lists, attempt limits, timeouts, output limits, the single
+  Codex emergency attempt, stage isolation and `fallback_to_host: never` are
+  unchanged by these preferences.
+- A requested model is prompt/config intent. Only the model reported in a
+  run's actual event envelope counts as effective; when the CLI does not report
+  it, record `not-observable`. Unknown context capacity, a missing agy login
+  or missing automatic compaction are recorded facts, not global fail-closed
+  conditions.
@@ -154,6 +154,30 @@ host 可以直接编写并提交的豁免只有 `.task/**`、`docs/superpowers/s
 - 会话隔离不要求为简单任务增加新平台或复杂编排。平凡文档/配置任务仍按最小闭环执行；需要独立 Review
  时只额外启动一个最小、一次性的评审 session，不复用实施 session，也不扩展任务范围。

+### 有界单元包、上下文与阶段交接
+
+通用规则见共享 [standards/agent-execution.md](../deps/development-guidelines/standards/agent-execution.md)
+与 [workflows/handoff.md](../deps/development-guidelines/workflows/handoff.md)；本节只写本仓做法，模型偏好见
+[AGENT-EXECUTION.md](../AGENT-EXECUTION.md#model-preferences-project-choice)。
+
+- **单元包**：用现有单元包 JSON 的 `goal`/`constraints`/`verification` 写清：一个结果及验收条件、允许路径与
+  须遵守的契约、非目标、前置提交 SHA、最小必读清单、直接验证命令、上下文容量及其来源（未知时按保守的
+  100K 假设）和为执行预留的余量。不新增 schema、运行时字段或控制器。
+- **容量**：100K/200K/300K/500K 只是规划参考，不是对任何 provider 能力的声明；容量更大不授权更宽的任务，
+  单元仍按一个结果拆分。容量未知不构成失败条件，也不做自动容量强制。
+- **阶段交接**：长任务每完成一个里程碑，host 自动在现有任务记录中写一份简短的持久交接：目标、状态、关键决定、
+  文件与 SHA、命令与结果、阻塞项、下一步和来源链接；下一个 CLI 阶段或单元以全新会话启动，只读取交接与最小
+  必读材料，不为交接等待人工。
+- **上下文压力**：provider 支持时，只在阶段完成后使用其上下文压缩；没有自动压缩不阻塞工作。普通上下文压力用
+  交接加 host 在既有授权内重新打包单元处理；只有真实扩围才按既有 `scope-expansion` fail-stop。
+- **如实区分**：请求模型与实际事件报告的有效模型；可观察的 session 身份与 `not-observable`；worker 局部自检
+  （SELF）与 host 集成验证、独立 Review、精确 HEAD CI（GATE）。
+- **验证分工**：worker 只跑直接检查；host 在最终候选上运行一次所需的集成套件。涉及安全、契约或部署的改动，
+  须写明正向、负向与回归用例。
+- **单写者与强制边界**：同一单元同一时刻只有一个写者，host 在运行期间不改规范检出。记录时区分三类约束：
+  提示词指导（单元包内容、容量、交接、模型偏好）、观察到的行为（事件报告的模型与 session），以及工具强制
+  （配置校验、尝试上限、超时、transport/admit）。本节不新增门禁，不放宽凭据、范围、未知副作用与 CI 边界。
+
 ### 两级分支交付模型与分流检查

 本仓采用两级业务交付模型：