@tea-agent/loop-agent 0.26.0 → 0.26.1

package/docs/templates/init-managed-agents.md

@@ -86,19 +86,24 @@ pwd → `README.md` → `harness.json` → `__LOOP_AGENT_GOVERNANCE_ROOT__/READM
 
 ```bash
 loop-agent new-task <task-id> "任务标题"
-# 有原始 PRD 时（推荐）：loop-agent import-prd <task-id> --file <path-to-prd.md>
+# 推荐默认：详细 PRD → import → prepare（无需手写两个 source；默认无 LLM 写 source）
+loop-agent import-prd <task-id> --file <path-to-prd.md>
 # 非微小或跨会话任务（推荐）：loop-agent plan create <plan-id> "<title>"
-# write .harness/tasks/<task-id>/source/需求.md
-# write .harness/tasks/<task-id>/source/执行约束.md
+loop-agent task source prepare <task-id> \
+  --use-imported-prd \
+  --allowed-path "<glob>" \
+  --forbidden-path ".harness/**" \
+  --verify "typecheck:npm run typecheck" \
+  --apply --json
 loop-agent dag run-task <task-id> --profile auto --strict-models
 loop-agent dag validate --dag .harness/tasks/<task-id>/dag.json --strict-models --strict-governance
 loop-agent run-dag --dag .harness/tasks/<task-id>/dag.json --cwd .
 # 有 plan 时收尾：loop-agent plan complete <plan-id> --summary "..."
 ```
 
-`source/需求.md` 与 `source/执行约束.md` 必需；有原始 PRD 时优先用 `import-prd` 原样归档。`import-prd` / `plan create` 不是 `dag run-task` 的硬依赖。写入前同步 `task.json.allowedPaths` / `task.json.forbiddenPaths` 并审查 writer `writeSet`。
+`source/需求.md` 与 `source/执行约束.md` 仍必需（M8/M9），但默认由 `task source prepare --apply` 投影生成，而不是主会话手写。有原始 PRD 时优先 `import-prd` 原样归档。`import-prd` / `plan create` 不是 `dag run-task` 的硬依赖。写入前同步 `task.json.allowedPaths` / `task.json.forbiddenPaths` 并审查 writer `writeSet`。
 
-凡是影响项目公共契约、执行入口、交付流水线、自动化/治理、数据模型、安全或权限模型、跨模块行为、用户可见工作流的改动，都必须在编辑实现文件前先创建任务、写好两个 source 文件、生成 DAG，并审查 DAG/writeSet。
+凡是影响项目公共契约、执行入口、交付流水线、自动化/治理、数据模型、安全或权限模型、跨模块行为、用户可见工作流的改动，都必须在编辑实现文件前先创建任务、完成 source prepare（或等价 managed contract）、生成 DAG，并审查 DAG/writeSet。
 
 ### 任务类型路由（taskKind）
 

package/docs/templates/interactive-ui-round2-experiment.md

@@ -1,66 +1,66 @@
-# Interactive UI Round-2 A/B/C Experiment
-
-## Frozen inputs
-
-- Target repo / commit:
-- Controller version:
-- TaskSpec / acceptance hash:
-- Base DAG:
-- Model matrix:
-
-Prepare the fixture and three DAGs after installing a published controller that contains `interactive-ui` support:
-
-```bash
-bash scripts-local/setup-drill-round2-react.sh /tmp/drill-round2-react-target
-npm run build
-node scripts-local/prepare-round2-ui-experiment.mjs \
-  /tmp/drill-round2-react-target \
-  features/F-2026-001/tasks/FE-001.yaml \
-  /tmp/round2-ui-experiment \
-  loop-agent
-node scripts-local/build-round2-ui-variants.mjs \
-  /tmp/round2-ui-experiment/round2-ui-base.json \
-  /tmp/round2-ui-experiment/variants
-```
-
-Validate every generated variant with the same frozen controller before running it. Use a unique run ID for A, B, and C; do not rewrite `executorModels`.
-
-## Variants
-
-| Variant | Writer prompt | Writer tier | Run ID | Result |
-| --- | --- | --- | --- | --- |
-| A | interactive-ui contract | MED | | |
-| B | default contract | HIGH | | |
-| C | interactive-ui contract | HIGH | | |
-
-## Metrics
-
-| Metric | A | B | C |
-| --- | --- | --- | --- |
-| First-pass review gate pass | | | |
-| Framework-native `.tsx` component | | | |
-| Route/parent integration | | | |
-| DOM interaction tests | | | |
-| Helper-only escape | | | |
-| Duration | | | |
-| Tokens | | | |
-
-## Required evidence per run
-
-- DAG JSON and run ID
-- implement/repair writer model and prompt profile
-- changed component path
-- integration path
-- interaction test path
-- DOM assertions mapped to AC-FE-001
-- review verdict and deterministic test output
-- diff boundary audit
-
-## Decision
-
-- Production default:
-- Evidence:
-- Cost/quality trade-off:
-- Follow-up:
-
-Do not conclude from a single run when provider or environment failures occurred. Re-run the affected variant with the same frozen inputs and a new run ID.
+# Interactive UI Round-2 A/B/C Experiment
+
+## Frozen inputs
+
+- Target repo / commit:
+- Controller version:
+- TaskSpec / acceptance hash:
+- Base DAG:
+- Model matrix:
+
+Prepare the fixture and three DAGs after installing a published controller that contains `interactive-ui` support:
+
+```bash
+bash scripts-local/setup-drill-round2-react.sh /tmp/drill-round2-react-target
+npm run build
+node scripts-local/prepare-round2-ui-experiment.mjs \
+  /tmp/drill-round2-react-target \
+  features/F-2026-001/tasks/FE-001.yaml \
+  /tmp/round2-ui-experiment \
+  loop-agent
+node scripts-local/build-round2-ui-variants.mjs \
+  /tmp/round2-ui-experiment/round2-ui-base.json \
+  /tmp/round2-ui-experiment/variants
+```
+
+Validate every generated variant with the same frozen controller before running it. Use a unique run ID for A, B, and C; do not rewrite `executorModels`.
+
+## Variants
+
+| Variant | Writer prompt | Writer tier | Run ID | Result |
+| --- | --- | --- | --- | --- |
+| A | interactive-ui contract | MED | | |
+| B | default contract | HIGH | | |
+| C | interactive-ui contract | HIGH | | |
+
+## Metrics
+
+| Metric | A | B | C |
+| --- | --- | --- | --- |
+| First-pass review gate pass | | | |
+| Framework-native `.tsx` component | | | |
+| Route/parent integration | | | |
+| DOM interaction tests | | | |
+| Helper-only escape | | | |
+| Duration | | | |
+| Tokens | | | |
+
+## Required evidence per run
+
+- DAG JSON and run ID
+- implement/repair writer model and prompt profile
+- changed component path
+- integration path
+- interaction test path
+- DOM assertions mapped to AC-FE-001
+- review verdict and deterministic test output
+- diff boundary audit
+
+## Decision
+
+- Production default:
+- Evidence:
+- Cost/quality trade-off:
+- Follow-up:
+
+Do not conclude from a single run when provider or environment failures occurred. Re-run the affected variant with the same frozen inputs and a new run ID.

package/docs/templates/knowledge-graph-bootstrap-dag.json

@@ -1,118 +1,118 @@
-{
-  "$schema": "./agent-dag.schema.json",
-  "version": 2,
-  "title": "Knowledge-graph bootstrap DAG template",
-  "objective": "AI-assisted business knowledge graph initialization: inventory repo signals, propose entities/edges under staging only, validate, review-gate, promote without overwrite, materialize query indexes.",
-  "successCriteria": [
-    "preflight requires knowledge/bootstrap/scope.yaml and status.yaml",
-    "inventory writes knowledge/bootstrap/inventory.json",
-    "propose-pi writes only knowledge/bootstrap/staging/** with evidence and non-asserted confidence",
-    "validate-shell fails on empty staging, missing edges.proposed/coverage-notes, or self-asserted confidence",
-    "multi-perspective review (structure/evidence/safety) each emit VERDICT; aggregate gate requires all pass before promote",
-    "promote merges new formal files only (no overwrite)",
-    "materialize writes knowledge/graph/entities-index.yaml and edges.yaml"
-  ],
-  "globalConstraints": [
-    "AI must never set confidence: asserted.",
-    "Do not modify src/** or Feature testing verdict/cases via bootstrap.",
-    "Promote must not overwrite existing formal knowledge files.",
-    "Complete B0/B1 skeleton before running this DAG."
-  ],
-  "defaults": {
-    "executor": "shell",
-    "contextProfile": "slim",
-    "writePolicy": "read-only"
-  },
-  "tasks": [
-    {
-      "id": "kg-bootstrap-preflight-shell",
-      "depends_on": [],
-      "role": "verifier",
-      "executor": "shell",
-      "complexity": "LOW",
-      "writePolicy": "read-only",
-      "subtask_prompt": "Require bootstrap skeleton."
-    },
-    {
-      "id": "kg-bootstrap-inventory-shell",
-      "depends_on": ["kg-bootstrap-preflight-shell"],
-      "role": "verifier",
-      "executor": "shell",
-      "complexity": "LOW",
-      "writePolicy": "exclusive",
-      "writeSet": ["knowledge/bootstrap/inventory.json"],
-      "subtask_prompt": "B2 inventory."
-    },
-    {
-      "id": "kg-bootstrap-propose-pi",
-      "depends_on": ["kg-bootstrap-inventory-shell"],
-      "role": "implementer",
-      "executor": "pi",
-      "toolProfile": "write",
-      "complexity": "HIGH",
-      "writePolicy": "exclusive",
-      "writeSet": [
-        "knowledge/bootstrap/staging/**",
-        "knowledge/bootstrap/runs/**"
-      ],
-      "subtask_prompt": "B3 AI proposals in staging only; never asserted."
-    },
-    {
-      "id": "kg-bootstrap-validate-shell",
-      "depends_on": ["kg-bootstrap-propose-pi"],
-      "role": "verifier",
-      "executor": "shell",
-      "complexity": "LOW",
-      "writePolicy": "read-only",
-      "subtask_prompt": "B4 validate staging."
-    },
-    {
-      "id": "kg-bootstrap-review-pi",
-      "depends_on": ["kg-bootstrap-validate-shell"],
-      "role": "reviewer",
-      "executor": "pi",
-      "complexity": "HIGH",
-      "writePolicy": "read-only",
-      "outputContract": "First line VERDICT: pass or VERDICT: request-revision.",
-      "subtask_prompt": "Review staging quality."
-    },
-    {
-      "id": "kg-bootstrap-review-gate-shell",
-      "depends_on": ["kg-bootstrap-review-pi"],
-      "role": "verifier",
-      "executor": "shell",
-      "complexity": "LOW",
-      "writePolicy": "read-only",
-      "subtask_prompt": "Gate on VERDICT: pass."
-    },
-    {
-      "id": "kg-bootstrap-promote-shell",
-      "depends_on": ["kg-bootstrap-review-gate-shell"],
-      "role": "verifier",
-      "executor": "shell",
-      "complexity": "LOW",
-      "writePolicy": "exclusive",
-      "writeSet": [
-        "knowledge/domains/**",
-        "knowledge/services/**",
-        "knowledge/modules/**",
-        "knowledge/graph/edges.manual.yaml",
-        "features/**/knowledge-links.yaml"
-      ],
-      "subtask_prompt": "B5 promote new files only."
-    },
-    {
-      "id": "kg-bootstrap-materialize-shell",
-      "depends_on": ["kg-bootstrap-promote-shell"],
-      "role": "closeout",
-      "executor": "shell",
-      "complexity": "LOW",
-      "writePolicy": "exclusive",
-      "writeSet": [
-        "knowledge/graph/entities-index.yaml",
-        "knowledge/graph/edges.yaml"
-      ],
-      "subtask_prompt": "B6 materialize indexes."
-    }
-  ]
-}
+{
+  "$schema": "./agent-dag.schema.json",
+  "version": 2,
+  "title": "Knowledge-graph bootstrap DAG template",
+  "objective": "AI-assisted business knowledge graph initialization: inventory repo signals, propose entities/edges under staging only, validate, review-gate, promote without overwrite, materialize query indexes.",
+  "successCriteria": [
+    "preflight requires knowledge/bootstrap/scope.yaml and status.yaml",
+    "inventory writes knowledge/bootstrap/inventory.json",
+    "propose-pi writes only knowledge/bootstrap/staging/** with evidence and non-asserted confidence",
+    "validate-shell fails on empty staging, missing edges.proposed/coverage-notes, or self-asserted confidence",
+    "multi-perspective review (structure/evidence/safety) each emit VERDICT; aggregate gate requires all pass before promote",
+    "promote merges new formal files only (no overwrite)",
+    "materialize writes knowledge/graph/entities-index.yaml and edges.yaml"
+  ],
+  "globalConstraints": [
+    "AI must never set confidence: asserted.",
+    "Do not modify src/** or Feature testing verdict/cases via bootstrap.",
+    "Promote must not overwrite existing formal knowledge files.",
+    "Complete B0/B1 skeleton before running this DAG."
+  ],
+  "defaults": {
+    "executor": "shell",
+    "contextProfile": "slim",
+    "writePolicy": "read-only"
+  },
+  "tasks": [
+    {
+      "id": "kg-bootstrap-preflight-shell",
+      "depends_on": [],
+      "role": "verifier",
+      "executor": "shell",
+      "complexity": "LOW",
+      "writePolicy": "read-only",
+      "subtask_prompt": "Require bootstrap skeleton."
+    },
+    {
+      "id": "kg-bootstrap-inventory-shell",
+      "depends_on": ["kg-bootstrap-preflight-shell"],
+      "role": "verifier",
+      "executor": "shell",
+      "complexity": "LOW",
+      "writePolicy": "exclusive",
+      "writeSet": ["knowledge/bootstrap/inventory.json"],
+      "subtask_prompt": "B2 inventory."
+    },
+    {
+      "id": "kg-bootstrap-propose-pi",
+      "depends_on": ["kg-bootstrap-inventory-shell"],
+      "role": "implementer",
+      "executor": "pi",
+      "toolProfile": "write",
+      "complexity": "HIGH",
+      "writePolicy": "exclusive",
+      "writeSet": [
+        "knowledge/bootstrap/staging/**",
+        "knowledge/bootstrap/runs/**"
+      ],
+      "subtask_prompt": "B3 AI proposals in staging only; never asserted."
+    },
+    {
+      "id": "kg-bootstrap-validate-shell",
+      "depends_on": ["kg-bootstrap-propose-pi"],
+      "role": "verifier",
+      "executor": "shell",
+      "complexity": "LOW",
+      "writePolicy": "read-only",
+      "subtask_prompt": "B4 validate staging."
+    },
+    {
+      "id": "kg-bootstrap-review-pi",
+      "depends_on": ["kg-bootstrap-validate-shell"],
+      "role": "reviewer",
+      "executor": "pi",
+      "complexity": "HIGH",
+      "writePolicy": "read-only",
+      "outputContract": "First line VERDICT: pass or VERDICT: request-revision.",
+      "subtask_prompt": "Review staging quality."
+    },
+    {
+      "id": "kg-bootstrap-review-gate-shell",
+      "depends_on": ["kg-bootstrap-review-pi"],
+      "role": "verifier",
+      "executor": "shell",
+      "complexity": "LOW",
+      "writePolicy": "read-only",
+      "subtask_prompt": "Gate on VERDICT: pass."
+    },
+    {
+      "id": "kg-bootstrap-promote-shell",
+      "depends_on": ["kg-bootstrap-review-gate-shell"],
+      "role": "verifier",
+      "executor": "shell",
+      "complexity": "LOW",
+      "writePolicy": "exclusive",
+      "writeSet": [
+        "knowledge/domains/**",
+        "knowledge/services/**",
+        "knowledge/modules/**",
+        "knowledge/graph/edges.manual.yaml",
+        "features/**/knowledge-links.yaml"
+      ],
+      "subtask_prompt": "B5 promote new files only."
+    },
+    {
+      "id": "kg-bootstrap-materialize-shell",
+      "depends_on": ["kg-bootstrap-promote-shell"],
+      "role": "closeout",
+      "executor": "shell",
+      "complexity": "LOW",
+      "writePolicy": "exclusive",
+      "writeSet": [
+        "knowledge/graph/entities-index.yaml",
+        "knowledge/graph/edges.yaml"
+      ],
+      "subtask_prompt": "B6 materialize indexes."
+    }
+  ]
+}