@tea-agent/loop-agent 0.13.0-beta.0 → 0.13.0

package/skills/loop-agent/references/harness-policy.md

@@ -1,263 +1,263 @@
-# Shared loop-agent Harness Policy
-
-本文件是跨仓库使用 `.` 的 canonical shared workflow policy。Repo-local harness docs 只应描述 local adapters：runtime 位置、governance root、适用的 verification commands。
-
-## Canonical stance
-
-- **Agent DAG** 是 medium/large、multi-file、architecture-sensitive、public-contract、CI/script 或 harness-runtime 工作的默认 implementation workflow。
-- 历史顺序式 `run analyze|plan|spec|implement|verify|auto|loop|continue` workflow 已移除。不要将其作为 fallback path 呈现。
-- **Long-running `loop`** 是 Agent DAG 之上的 outer state/evidence layer。它记录 rounds、context compression、signals、canonical refs；不得替代 complex work 的 DAG writeSet review、Decision Gate 或 shell verification。
-- **Main session** 负责 orchestrate：选一个 work chunk、准备 source materials、review DAG/writeSet、monitor failures、跑 final verification、hand off。
-- **Executors** 实现 bounded work：Pi 是唯一受治理 Agent runtime（read-only planning/review/diagnosis，以及 `toolProfile: "write"` 的 bounded implementation/repair）；shell 产出 deterministic verification facts；`cursor-prompt` 仅是 one-shot sidecar。
-- **Shell verification 是 completion fact source**。LLM review 或 advisory output 不能替代 command exit codes 与 archived evidence。
-
-## Command surface tiers
-
-| Tier | Default purpose | Commands |
-|---|---|---|
-| Primary | Normal autonomous implementation | `new-task` -> `dag run-task --profile auto` -> `dag validate --strict-models --strict-governance` -> `run-dag` |
-| Operator | Diagnose, recover, close out, inspect facts | `status`, `instructions`, `dag status`, `dag doctor`, `dag report`, `dag closeout-draft`, `dag reconcile-tasks`, `dag final-verification`, `inspect`, `doctor`, `spine audit`, `knowledge curate`, `docs audit`, `handoff check`, `loop-benchmark` |
-| Compatibility | Legacy task metadata and feature-study helpers | `goal`, `reference`, `study` |
-| Escape hatch | Isolated delegation, one-shot diagnosis or bounded repair | `delegate`, `worktree`, `harvest`, `pi-prompt`, `cursor-prompt` |
-| Experimental | Long-running outer task state | `loop init|status|run|record-round|add-signal|closeout` |
-
-Prompt templates、README snippets、task instructions 应优先呈现 Primary + Operator。Compatibility 与 escape-hatch commands 仍可用，但须携带其 downgrade/fallback 含义。
-
-## Entry selection decision tree
-
-```text
-Is this only status, diagnosis, recovery, or closeout?
-  yes -> Operator commands.
-  no  -> Does it need recoverable, reviewable, verifiable implementation state?
-           no  -> Use read-only pi-prompt for analysis, or a tiny main-session surgical patch only if obvious and immediately verifiable.
-           yes -> Agent DAG.
-```
-
-在以下任一 signal 适用时用 Agent DAG 而非 broad one-shot execution：
-
-- loop-agent runtime, DAG schema, run facts, promotion/closeout, scripts/CI, public contract, or shared protocol is touched.
-- The change needs multiple files, multiple scouts, review gates, Decision Gate, repair flow, or shell gate.
-- `writeSet` is broad, multiple exclusive writers exist, or public interfaces / architecture boundaries change.
-- Requirement, architecture, credential, cost, deployment, security, or authority surface is unclear.
-- A failure repeats and needs recovery planning rather than blind retry.
-
-## Agent DAG path
-
-Minimum governed path：
-
-```bash
-loop-agent new-task <task-id> "Task Title" [--repo-root <target-repo>]
-# optional but recommended for user PRDs:
-# loop-agent import-prd <task-id> --file <path-to-original-prd.md> [--repo-root <target-repo>]
-# write derived <target-repo>/.harness/tasks/<task-id>/source/需求.md
-# write <target-repo>/.harness/tasks/<task-id>/source/执行约束.md
-
-loop-agent dag run-task <task-id> \
-  --profile auto \
-  --strict-models \
-  --output <temp-dir>/<task-id>-dag.json \
-  [--repo-root <target-repo>]
-
-loop-agent dag validate \
-  --dag <temp-dir>/<task-id>-dag.json \
-  --strict-models \
-  --strict-governance
-
-loop-agent run-dag \
-  --dag <temp-dir>/<task-id>-dag.json \
-  --cwd <target-repo>
-```
-
-`loop-agent` is the preferred global CLI. For self-hosting loop-agent development, the controller must be an installed npm-published package. Use `npm install -g @tea-agent/loop-agent@latest` for first install or intentional upgrades, then treat the installed version as frozen for the current task and record `npm list -g @tea-agent/loop-agent --depth=0`. Do not repeatedly fetch `npx @latest` inside DAG nodes, and do not use the current working tree's `npm link` or `npm run dev` to control tasks that may edit CLI, DAG runtime, executors, package metadata, or build output. Use `npm run dev -- <args>` only for source debugging and focused CLI development.
-
-The npm package carries static capability assets: `skills/`, top-level governance docs, `docs/templates/`, `examples/`, `harness.json`, `AGENTS.md`, `README.md`, and `CHANGELOG.md`. Generated or historical task facts under `docs/progress/`, `docs/reports/`, `docs/exec-plans/`, and `docs/decisions/` belong to the target repository; package only their directory README files, not prior run content.
-
-For arbitrary target repositories, DAG skill instructions must not depend on loop-agent source history being copied into the target repo. Resolve configured, user, or target-local skills when present, then fall back to package-bundled `skills/` as the stable default capability set.
-
-`<temp-dir>` means the platform-native temp directory. Use native paths for actual `--output`, `--dag`, and `--cwd` values on macOS and Windows; use `/` only for stable repo refs, JSON/Markdown evidence refs, and glob conventions.
-
-Execution 前 review `dag run-task` JSON / `reviewPacket`：
-
-- `profileRouting`：requested profile、selected profile/template、routing reasons。
-- `governanceProfile`：process、delivery、code-change signals。
-- Writer nodes：`writePolicy`、`writeSet`、`allowedPaths`、`forbiddenPaths`、broad entries、forbidden overlaps。
-- Shell gates 与 verification commands。
-- Decision Gate mode（`record-only` vs `pause-on-human`）。
-- 执行前须 narrow 的 placeholder、`**` 或 repo-root writeSet。
-
-In-flight DAG shell checks 在需要时用 repo active-run override（例如 `HARNESS_ALLOW_ACTIVE_DAG_RUNS=1 bash scripts/check-repo.sh`）。DAG archived 后，再不带 in-flight override 跑 repo check。
-
-On Windows, run Bash scripts through Git Bash or a configured compatible Bash. Do not require WSL, `/tmp`, `which`, or other POSIX filesystem assumptions in loop-agent CLI behavior.
-
-## Task source materials
-
-每个 handoff-ready task 包含：
-
-```text
-.harness/tasks/<task-id>/
-  task.json
-  source/
-    references/          # immutable original PRD / acceptance / design
-    source-manifest.json # optional hash manifest from import-prd
-    需求.md              # derived execution contract
-    执行约束.md
-```
-
-`需求.md` 应陈述 objective、scope、non-goals、acceptance criteria，并链接或追溯 `source/references/*` / repo-local specs 或 plans。原始 PRD 优先 `import-prd` 归档；Worker materialize 会复制 `source_docs` 到 `references/`，并把 acceptance_refs 展开为短摘要。冲突时以 `source/references/*` 为准。
-
-`执行约束.md` 应陈述：
-
-- allowed paths
-- forbidden paths
-- 当前 dirty workspace / protected user changes（如有）
-- architecture boundaries 与 invariants
-- expected verification commands
-- acceptance criteria / failure conditions
-- 是否允许 DAG fallback，及若已知时的 fallback reason
-
-若 `spec`、`plan` 或 DAG generation 后 source materials 变更，implementation 前 regenerate 或 revalidate plan/DAG。
-
-## Long-running loop policy
-
-`loop` 用于 long-running outer task memory：objective/context projection、round records、signals、derived events、verification summaries、closeout draft。它不是 Agent DAG 的 substitute。
-
-Governed work 的典型 loop path：
-
-```bash
-loop-agent loop init <task-id>
-loop-agent loop run <task-id> --action dag
-# review DAG packet / writeSet / shell gates
-loop-agent loop run <task-id> --action dag --execute
-loop-agent loop run <task-id> --action shell-verify --command "<repo-check>"
-loop-agent loop run <task-id> --action pi-review
-loop-agent loop run <task-id> --auto --max-rounds 3
-loop-agent loop closeout <task-id>
-```
-
-Loop action rules：
-
-- `shell-verify` 是 deterministic；exit code 决定 verification record。
-- `pi-review` 是 read-only；tools 限于 `read,grep,find,ls`，output 为 structured advisory evidence。Structured JSON 须含 `findingSummary`、`failureCategory`、`nextHypothesis`、`recommendedAction`、`fixScope`、`rootCause`；`recommendedAction`  exactly 为 `implement_fix|replan|pause|done`。
-- 自动写入只能通过 Pi-only Agent DAG execute；须读 task `allowedPaths` / `forbiddenPaths`，审查 writer `writeSet`，并 follow shell verification 或 review。
-- 对 `task.json.complexity = medium | large`，自动 DAG execute  additionally 需要：
-  - previous loop `dag` round，或
-  - explicit `task.json.dagFallbackReason` 说明为何不能用 DAG。
-- `loop run --auto` 默认不 write。Auto DAG execute 需要 `task.json.loopAutoExecutionPolicy="enabled"`，或 `loopAutoExecutionPolicy="approval-required"` 加 pending approval signal；旧 `loopAutoWritePolicy` fail-fast；write guards 仍 fail closed 并 pause。
-- `loop closeout` 须报告 workflow path：`dag`、`explicit-fallback`、`missing-dag-evidence` 或 `micro-or-small`。
-- 无 DAG evidence 且无 `dagFallbackReason` 的 medium/large closeout 须将其列为 remaining risk。
-- `record-round --decision complete` 仅是 loop-state candidate；completion 仍须 shell verification、review verdict、success-criteria coverage。
-
-## Supervised DAG convergence
-
-Supervised DAG convergence 可选且由 task-config 驱动：
-
-```json
-{
-  "convergence": {
-    "enabled": true,
-    "maxPasses": 3,
-    "stopOnHardVerifyPass": true,
-    "pauseOnRegression": true
-  }
-}
-```
-
-Rules:
-
-- 默认保持 single repair，除非 `convergence.enabled=true`；`HARNESS_DAG_CONVERGENCE=off` 是 rollback switch。
-- Supervised process supervisor 须 emit 首行 `VERDICT:` 与 `REPAIR_ARTIFACT_JSON` fenced block。Repair prompts 应先消费 artifact `failureClass`、`rootCause`、`fixScope`、`invariant`；raw logs 仅在 artifact 允许时为 fallback evidence。
-- 在 `maxPasses` 前 retryable `hard-verify-shell` failure 时，preserve current pass evidence 于 `convergence/pass-N/`，reset process-supervisor/process-gate/repair/hard-verify segment 及 blocked downstream nodes，再进入现有 DAG rank execution loop。
-- 不要 retry write guards、timeout/spawn/auth failures 或 human-gate failures。
-- 出现 conservative regression signals（如 lower shell success count）时 pause 而非 retry。
-- `dag report --json` 与 markdown 须 expose `convergence.passHistory`。
-- Final completion authority 仍是 full shell verification；quota/focused commands 仅为 intermediate cost controls。
-
-## Structured repair, spine audit, and curator gates
-
-- `shell.repairArtifactGate.fromNodeId` validates the upstream supervisor artifact before repair. Missing/invalid JSON, missing request-revision `fixScope`, or scope outside the downstream repair writer allowedPaths/writeSet fails closed.
-- `spine audit <task-id>` is the deterministic minimal spec spine checker for task source, ownership paths, requirement coverage, and final verification commands.
-- `dag validate --strict-governance --spine-task <task-id>` may consume the same spine audit as part of strict validation.
-- `knowledge curate` reads completed convergence patterns and writes only human-gated proposal Markdown after skill safety preflight.
-
-## SePO-lite prompt evolution
-
-- Learned prompt deltas 是 human-gated proposals；成为 reusable guidance 前须 review。
-- Prompt deltas 为 Markdown-only process guidance；不得含 shell commands、credential handling、tool permission expansion 或 completion-authority bypass。
-- Accepted learned guidance 位于 `./skill/references/learned/<repo>.md` 或 `default.md`。
-- 已 request `loop-agent` 的 DAG implementer prompts 可 inline 最多三个 human-gated learned Markdown sections。
-- Learned guidance 为 advisory，永不替代 writeSet governance、Decision Gate policy 或 shell verification。
-
-## Sidecar interventions
-
-`pi-prompt` 与 `cursor-prompt` 是 sidecar interventions，不是 workflow state。
-
-用 `pi-prompt` 做短时 read-only planning、log explanation 或 failure diagnosis。Read-only 时传 read-only tools 并写明不 edit files：
-
-```bash
-loop-agent pi-prompt \
-  --cwd <repo-root> \
-  --tools read,grep,find,ls \
-  --timeout 2400000 \
-  "Read the task source and diagnose the failure. Do not edit files."
-```
-
-用 `cursor-prompt` 做 bounded multi-file diagnosis 或 small repair，prompt 须含：
-
-- task id
-- exact objective
-- allowed paths
-- forbidden paths
-- hard constraints
-- expected verification
-- instruction to preserve unrelated files
-
-Sidecar output 为 advisory。若须成为 task evidence，通过 loop-agent run/task artifacts promote 或 summarize；completed DAG 与 one-shot run facts 保持只读。
-
-## Model and executor boundaries
-
-- Agent DAG 用 DAG JSON `executorModels` 加 node `executor` / `complexity`；不要从 repo `harness.json.models` 推断 DAG models。
-- DAG `shell` 与 `static` nodes 不用 models。
-- `harness.json.models.<step>` 下 historical step models 是 legacy metadata，不是新 DAG work 的 routing。
-- `pi-prompt` / `cursor-prompt` models 来自 CLI flags 或 runtime defaults，须 per intervention 选择。
-- Pi DAG nodes 默认 read-only planning/review/diagnosis；声明 `toolProfile: "write"` 时是 bounded writers，须有 explicit write scope。
-- 受治理 writer 固定为 `implement-pi` / `repair-pi`，须有 explicit write scope；`cursor-prompt` 不进入 DAG schema。
-- Shell nodes 产出 deterministic verification facts 与 gates。
-
-## Artifacts and facts boundary
-
-- `.harness/tasks/<task-id>/` 是 task runtime state。
-- `.harness/tasks/<task-id>/loop/` 是 loop runtime projection；不替代 task source 或 repo specs。
-- `.harness/dag-runs/{active,paused,completed}/<run-id>/` 是 DAG run fact storage。Completed facts 为 read-only。
-- `.harness/runs/{active,completed,failed}/<run-id>/` 是 one-shot Pi/Cursor evidence。Completed/failed facts 为 read-only。
-- `.harness/task-pool/` 是伴生 CLI `agent-worker`（Worker TaskSpec pipeline）的 runtime state：batch artifacts、Task Pool JSONL/state、晨报和 failure handoffs。默认被忽略，不提交。
-- Root `artifacts/` 是 legacy/current-work summary space，不是 DAG read-only scratchpad，也不是新 DAG work 的 default handoff。
-- Long-term conclusions 属于 repo governance docs、progress、reports、decisions、tests 或 scripts。
-
-除非 task 显式 promote trimmed report 到 repo governance docs，不要提交 `.harness/dag-runs/`、`.harness/runs/`、`.harness/cache/` 或 `.harness/task-pool/` 的 runtime histories。
-
-## Baseline, dirty workspace, and verification
-
-Complex implementation 前：
-
-1. Check current directory 与 target repo。
-2. Read repo entrypoints（`README`、`AGENTS`、`harness.json`、governance index）。
-3. Capture affected area 的 minimal baseline verification。
-4. 若 workspace dirty，选一：
-   - isolated worktree，或
-   - explicit user confirmation 在当前 workspace 工作并 preserve/possibly include existing changes。
-5. Record known baseline failures，足以区分 pre-existing failures 与 task regressions。
-
-Verification 应从 target repo verification matrix 选择。Cross-repo documentation refactors 时在 each affected repo 跑 checks。
-
-## Handoff requirements
-
-每个 task handoff 应回答：
-
-1. What changed and why。
-2. 用了哪条 workflow path：DAG、sidecar 或 main-session surgical patch。
-3. 若从 DAG downgrade，explicit reason 与 evidence。
-4. Executors used 及其 boundaries。
-5. Verification commands run 与 results。
-6. DAG / one-shot / loop refs（如有）。
-7. Remaining risks 与 follow-up tasks。
-8. 是否应将 new rules promote 到 docs、tests、scripts 或 shared skill references。
+# Shared loop-agent Harness Policy
+
+本文件是跨仓库使用 `.` 的 canonical shared workflow policy。Repo-local harness docs 只应描述 local adapters：runtime 位置、governance root、适用的 verification commands。
+
+## Canonical stance
+
+- **Agent DAG** 是 medium/large、multi-file、architecture-sensitive、public-contract、CI/script 或 harness-runtime 工作的默认 implementation workflow。
+- 历史顺序式 `run analyze|plan|spec|implement|verify|auto|loop|continue` workflow 已移除。不要将其作为 fallback path 呈现。
+- **Long-running `loop`** 是 Agent DAG 之上的 outer state/evidence layer。它记录 rounds、context compression、signals、canonical refs；不得替代 complex work 的 DAG writeSet review、Decision Gate 或 shell verification。
+- **Main session** 负责 orchestrate：选一个 work chunk、准备 source materials、review DAG/writeSet、monitor failures、跑 final verification、hand off。
+- **Executors** 实现 bounded work：Pi 是唯一受治理 Agent runtime（read-only planning/review/diagnosis，以及 `toolProfile: "write"` 的 bounded implementation/repair）；shell 产出 deterministic verification facts；`cursor-prompt` 仅是 one-shot sidecar。
+- **Shell verification 是 completion fact source**。LLM review 或 advisory output 不能替代 command exit codes 与 archived evidence。
+
+## Command surface tiers
+
+| Tier | Default purpose | Commands |
+|---|---|---|
+| Primary | Normal autonomous implementation | `new-task` -> `dag run-task --profile auto` -> `dag validate --strict-models --strict-governance` -> `run-dag` |
+| Operator | Diagnose, recover, close out, inspect facts | `status`, `instructions`, `dag status`, `dag doctor`, `dag report`, `dag closeout-draft`, `dag reconcile-tasks`, `dag final-verification`, `inspect`, `doctor`, `spine audit`, `knowledge curate`, `docs audit`, `handoff check`, `loop-benchmark` |
+| Compatibility | Legacy task metadata and feature-study helpers | `goal`, `reference`, `study` |
+| Escape hatch | Isolated delegation, one-shot diagnosis or bounded repair | `delegate`, `worktree`, `harvest`, `pi-prompt`, `cursor-prompt` |
+| Experimental | Long-running outer task state | `loop init|status|run|record-round|add-signal|closeout` |
+
+Prompt templates、README snippets、task instructions 应优先呈现 Primary + Operator。Compatibility 与 escape-hatch commands 仍可用，但须携带其 downgrade/fallback 含义。
+
+## Entry selection decision tree
+
+```text
+Is this only status, diagnosis, recovery, or closeout?
+  yes -> Operator commands.
+  no  -> Does it need recoverable, reviewable, verifiable implementation state?
+           no  -> Use read-only pi-prompt for analysis, or a tiny main-session surgical patch only if obvious and immediately verifiable.
+           yes -> Agent DAG.
+```
+
+在以下任一 signal 适用时用 Agent DAG 而非 broad one-shot execution：
+
+- loop-agent runtime, DAG schema, run facts, promotion/closeout, scripts/CI, public contract, or shared protocol is touched.
+- The change needs multiple files, multiple scouts, review gates, Decision Gate, repair flow, or shell gate.
+- `writeSet` is broad, multiple exclusive writers exist, or public interfaces / architecture boundaries change.
+- Requirement, architecture, credential, cost, deployment, security, or authority surface is unclear.
+- A failure repeats and needs recovery planning rather than blind retry.
+
+## Agent DAG path
+
+Minimum governed path：
+
+```bash
+loop-agent new-task <task-id> "Task Title" [--repo-root <target-repo>]
+# optional but recommended for user PRDs:
+# loop-agent import-prd <task-id> --file <path-to-original-prd.md> [--repo-root <target-repo>]
+# write derived <target-repo>/.harness/tasks/<task-id>/source/需求.md
+# write <target-repo>/.harness/tasks/<task-id>/source/执行约束.md
+
+loop-agent dag run-task <task-id> \
+  --profile auto \
+  --strict-models \
+  --output <temp-dir>/<task-id>-dag.json \
+  [--repo-root <target-repo>]
+
+loop-agent dag validate \
+  --dag <temp-dir>/<task-id>-dag.json \
+  --strict-models \
+  --strict-governance
+
+loop-agent run-dag \
+  --dag <temp-dir>/<task-id>-dag.json \
+  --cwd <target-repo>
+```
+
+`loop-agent` is the preferred global CLI. For self-hosting loop-agent development, the controller must be an installed npm-published package. Use `npm install -g @tea-agent/loop-agent@latest` for first install or intentional upgrades, then treat the installed version as frozen for the current task and record `npm list -g @tea-agent/loop-agent --depth=0`. Do not repeatedly fetch `npx @latest` inside DAG nodes, and do not use the current working tree's `npm link` or `npm run dev` to control tasks that may edit CLI, DAG runtime, executors, package metadata, or build output. Use `npm run dev -- <args>` only for source debugging and focused CLI development.
+
+The npm package carries static capability assets: `.agents/skills/`, top-level governance docs, `ai_workspace/loop-agent/templates/`, `examples/`, `harness.json`, `AGENTS.md`, `README.md`, and `CHANGELOG.md`. Generated or historical task facts under `ai_workspace/loop-agent/progress/`, `ai_workspace/loop-agent/reports/`, `ai_workspace/loop-agent/exec-plans/`, and `ai_workspace/loop-agent/decisions/` belong to the target repository; package only their directory README files, not prior run content.
+
+For arbitrary target repositories, DAG skill instructions must not depend on loop-agent source history being copied into the target repo. Resolve configured, user, or target-local skills when present, then fall back to package-bundled `.agents/skills/` as the stable default capability set.
+
+`<temp-dir>` means the platform-native temp directory. Use native paths for actual `--output`, `--dag`, and `--cwd` values on macOS and Windows; use `/` only for stable repo refs, JSON/Markdown evidence refs, and glob conventions.
+
+Execution 前 review `dag run-task` JSON / `reviewPacket`：
+
+- `profileRouting`：requested profile、selected profile/template、routing reasons。
+- `governanceProfile`：process、delivery、code-change signals。
+- Writer nodes：`writePolicy`、`writeSet`、`allowedPaths`、`forbiddenPaths`、broad entries、forbidden overlaps。
+- Shell gates 与 verification commands。
+- Decision Gate mode（`record-only` vs `pause-on-human`）。
+- 执行前须 narrow 的 placeholder、`**` 或 repo-root writeSet。
+
+In-flight DAG shell checks 在需要时用 repo active-run override（例如 `HARNESS_ALLOW_ACTIVE_DAG_RUNS=1 bash scripts/check-repo.sh`）。DAG archived 后，再不带 in-flight override 跑 repo check。
+
+On Windows, run Bash scripts through Git Bash or a configured compatible Bash. Do not require WSL, `/tmp`, `which`, or other POSIX filesystem assumptions in loop-agent CLI behavior.
+
+## Task source materials
+
+每个 handoff-ready task 包含：
+
+```text
+.harness/tasks/<task-id>/
+  task.json
+  source/
+    references/          # immutable original PRD / acceptance / design
+    source-manifest.json # optional hash manifest from import-prd
+    需求.md              # derived execution contract
+    执行约束.md
+```
+
+`需求.md` 应陈述 objective、scope、non-goals、acceptance criteria，并链接或追溯 `source/references/*` / repo-local specs 或 plans。原始 PRD 优先 `import-prd` 归档；Worker materialize 会复制 `source_docs` 到 `references/`，并把 acceptance_refs 展开为短摘要。冲突时以 `source/references/*` 为准。
+
+`执行约束.md` 应陈述：
+
+- allowed paths
+- forbidden paths
+- 当前 dirty workspace / protected user changes（如有）
+- architecture boundaries 与 invariants
+- expected verification commands
+- acceptance criteria / failure conditions
+- 是否允许 DAG fallback，及若已知时的 fallback reason
+
+若 `spec`、`plan` 或 DAG generation 后 source materials 变更，implementation 前 regenerate 或 revalidate plan/DAG。
+
+## Long-running loop policy
+
+`loop` 用于 long-running outer task memory：objective/context projection、round records、signals、derived events、verification summaries、closeout draft。它不是 Agent DAG 的 substitute。
+
+Governed work 的典型 loop path：
+
+```bash
+loop-agent loop init <task-id>
+loop-agent loop run <task-id> --action dag
+# review DAG packet / writeSet / shell gates
+loop-agent loop run <task-id> --action dag --execute
+loop-agent loop run <task-id> --action shell-verify --command "<repo-check>"
+loop-agent loop run <task-id> --action pi-review
+loop-agent loop run <task-id> --auto --max-rounds 3
+loop-agent loop closeout <task-id>
+```
+
+Loop action rules：
+
+- `shell-verify` 是 deterministic；exit code 决定 verification record。
+- `pi-review` 是 read-only；tools 限于 `read,grep,find,ls`，output 为 structured advisory evidence。Structured JSON 须含 `findingSummary`、`failureCategory`、`nextHypothesis`、`recommendedAction`、`fixScope`、`rootCause`；`recommendedAction`  exactly 为 `implement_fix|replan|pause|done`。
+- 自动写入只能通过 Pi-only Agent DAG execute；须读 task `allowedPaths` / `forbiddenPaths`，审查 writer `writeSet`，并 follow shell verification 或 review。
+- 对 `task.json.complexity = medium | large`，自动 DAG execute  additionally 需要：
+  - previous loop `dag` round，或
+  - explicit `task.json.dagFallbackReason` 说明为何不能用 DAG。
+- `loop run --auto` 默认不 write。Auto DAG execute 需要 `task.json.loopAutoExecutionPolicy="enabled"`，或 `loopAutoExecutionPolicy="approval-required"` 加 pending approval signal；旧 `loopAutoWritePolicy` fail-fast；write guards 仍 fail closed 并 pause。
+- `loop closeout` 须报告 workflow path：`dag`、`explicit-fallback`、`missing-dag-evidence` 或 `micro-or-small`。
+- 无 DAG evidence 且无 `dagFallbackReason` 的 medium/large closeout 须将其列为 remaining risk。
+- `record-round --decision complete` 仅是 loop-state candidate；completion 仍须 shell verification、review verdict、success-criteria coverage。
+
+## Supervised DAG convergence
+
+Supervised DAG convergence 可选且由 task-config 驱动：
+
+```json
+{
+  "convergence": {
+    "enabled": true,
+    "maxPasses": 3,
+    "stopOnHardVerifyPass": true,
+    "pauseOnRegression": true
+  }
+}
+```
+
+Rules:
+
+- 默认保持 single repair，除非 `convergence.enabled=true`；`HARNESS_DAG_CONVERGENCE=off` 是 rollback switch。
+- Supervised process supervisor 须 emit 首行 `VERDICT:` 与 `REPAIR_ARTIFACT_JSON` fenced block。Repair prompts 应先消费 artifact `failureClass`、`rootCause`、`fixScope`、`invariant`；raw logs 仅在 artifact 允许时为 fallback evidence。
+- 在 `maxPasses` 前 retryable `hard-verify-shell` failure 时，preserve current pass evidence 于 `convergence/pass-N/`，reset process-supervisor/process-gate/repair/hard-verify segment 及 blocked downstream nodes，再进入现有 DAG rank execution loop。
+- 不要 retry write guards、timeout/spawn/auth failures 或 human-gate failures。
+- 出现 conservative regression signals（如 lower shell success count）时 pause 而非 retry。
+- `dag report --json` 与 markdown 须 expose `convergence.passHistory`。
+- Final completion authority 仍是 full shell verification；quota/focused commands 仅为 intermediate cost controls。
+
+## Structured repair, spine audit, and curator gates
+
+- `shell.repairArtifactGate.fromNodeId` validates the upstream supervisor artifact before repair. Missing/invalid JSON, missing request-revision `fixScope`, or scope outside the downstream repair writer allowedPaths/writeSet fails closed.
+- `spine audit <task-id>` is the deterministic minimal spec spine checker for task source, ownership paths, requirement coverage, and final verification commands.
+- `dag validate --strict-governance --spine-task <task-id>` may consume the same spine audit as part of strict validation.
+- `knowledge curate` reads completed convergence patterns and writes only human-gated proposal Markdown after skill safety preflight.
+
+## SePO-lite prompt evolution
+
+- Learned prompt deltas 是 human-gated proposals；成为 reusable guidance 前须 review。
+- Prompt deltas 为 Markdown-only process guidance；不得含 shell commands、credential handling、tool permission expansion 或 completion-authority bypass。
+- Accepted learned guidance 位于 `./skill/references/learned/<repo>.md` 或 `default.md`。
+- 已 request `loop-agent` 的 DAG implementer prompts 可 inline 最多三个 human-gated learned Markdown sections。
+- Learned guidance 为 advisory，永不替代 writeSet governance、Decision Gate policy 或 shell verification。
+
+## Sidecar interventions
+
+`pi-prompt` 与 `cursor-prompt` 是 sidecar interventions，不是 workflow state。
+
+用 `pi-prompt` 做短时 read-only planning、log explanation 或 failure diagnosis。Read-only 时传 read-only tools 并写明不 edit files：
+
+```bash
+loop-agent pi-prompt \
+  --cwd <repo-root> \
+  --tools read,grep,find,ls \
+  --timeout 2400000 \
+  "Read the task source and diagnose the failure. Do not edit files."
+```
+
+用 `cursor-prompt` 做 bounded multi-file diagnosis 或 small repair，prompt 须含：
+
+- task id
+- exact objective
+- allowed paths
+- forbidden paths
+- hard constraints
+- expected verification
+- instruction to preserve unrelated files
+
+Sidecar output 为 advisory。若须成为 task evidence，通过 loop-agent run/task artifacts promote 或 summarize；completed DAG 与 one-shot run facts 保持只读。
+
+## Model and executor boundaries
+
+- Agent DAG 用 DAG JSON `executorModels` 加 node `executor` / `complexity`；不要从 repo `harness.json.models` 推断 DAG models。
+- DAG `shell` 与 `static` nodes 不用 models。
+- `harness.json.models.<step>` 下 historical step models 是 legacy metadata，不是新 DAG work 的 routing。
+- `pi-prompt` / `cursor-prompt` models 来自 CLI flags 或 runtime defaults，须 per intervention 选择。
+- Pi DAG nodes 默认 read-only planning/review/diagnosis；声明 `toolProfile: "write"` 时是 bounded writers，须有 explicit write scope。
+- 受治理 writer 固定为 `implement-pi` / `repair-pi`，须有 explicit write scope；`cursor-prompt` 不进入 DAG schema。
+- Shell nodes 产出 deterministic verification facts 与 gates。
+
+## Artifacts and facts boundary
+
+- `.harness/tasks/<task-id>/` 是 task runtime state。
+- `.harness/tasks/<task-id>/loop/` 是 loop runtime projection；不替代 task source 或 repo specs。
+- `.harness/dag-runs/{active,paused,completed}/<run-id>/` 是 DAG run fact storage。Completed facts 为 read-only。
+- `.harness/runs/{active,completed,failed}/<run-id>/` 是 one-shot Pi/Cursor evidence。Completed/failed facts 为 read-only。
+- `.harness/task-pool/` 是伴生 CLI `agent-worker`（Worker TaskSpec pipeline）的 runtime state：batch artifacts、Task Pool JSONL/state、晨报和 failure handoffs。默认被忽略，不提交。
+- Root `artifacts/` 是 legacy/current-work summary space，不是 DAG read-only scratchpad，也不是新 DAG work 的 default handoff。
+- Long-term conclusions 属于 repo governance docs、progress、reports、decisions、tests 或 scripts。
+
+除非 task 显式 promote trimmed report 到 repo governance docs，不要提交 `.harness/dag-runs/`、`.harness/runs/`、`.harness/cache/` 或 `.harness/task-pool/` 的 runtime histories。
+
+## Baseline, dirty workspace, and verification
+
+Complex implementation 前：
+
+1. Check current directory 与 target repo。
+2. Read repo entrypoints（`README`、`AGENTS`、`harness.json`、governance index）。
+3. Capture affected area 的 minimal baseline verification。
+4. 若 workspace dirty，选一：
+   - isolated worktree，或
+   - explicit user confirmation 在当前 workspace 工作并 preserve/possibly include existing changes。
+5. Record known baseline failures，足以区分 pre-existing failures 与 task regressions。
+
+Verification 应从 target repo verification matrix 选择。Cross-repo documentation refactors 时在 each affected repo 跑 checks。
+
+## Handoff requirements
+
+每个 task handoff 应回答：
+
+1. What changed and why。
+2. 用了哪条 workflow path：DAG、sidecar 或 main-session surgical patch。
+3. 若从 DAG downgrade，explicit reason 与 evidence。
+4. Executors used 及其 boundaries。
+5. Verification commands run 与 results。
+6. DAG / one-shot / loop refs（如有）。
+7. Remaining risks 与 follow-up tasks。
+8. 是否应将 new rules promote 到 docs、tests、scripts 或 shared skill references。