@tea-agent/loop-agent 0.15.0 → 0.16.0

package/docs/init-surface.manifest.json

@@ -39,6 +39,7 @@
     "docs/templates/product-line/test-plan.md",
     "docs/templates/product-line/task-graph.yaml",
     "docs/templates/product-line/task.yaml",
+    "docs/templates/product-line/feature.yaml",
     "docs/templates/product-line/closeout.yaml",
     "docs/templates/product-line/links.md",
     "scripts/kb-bootstrap-init-skeleton.sh",
@@ -106,6 +107,7 @@
     "docs/templates/product-line/test-plan.md",
     "docs/templates/product-line/task-graph.yaml",
     "docs/templates/product-line/task.yaml",
+    "docs/templates/product-line/feature.yaml",
     "docs/templates/product-line/closeout.yaml",
     "docs/templates/product-line/links.md",
     "scripts/check-skill-entry.sh",
@@ -175,6 +177,7 @@
     "docs/templates/product-line/test-plan.md": "copied",
     "docs/templates/product-line/task-graph.yaml": "copied",
     "docs/templates/product-line/task.yaml": "copied",
+    "docs/templates/product-line/feature.yaml": "copied",
     "docs/templates/product-line/closeout.yaml": "copied",
     "docs/templates/product-line/links.md": "copied",
     "scripts/check-skill-entry.sh": "generated",

package/docs/templates/evaluation/campaign-budget-v1.json

@@ -0,0 +1,12 @@
+{
+	"schemaVersion": 1,
+	"mode": "hard",
+	"limits": {
+		"maxTokens": 2000000,
+		"maxWallTimeMs": 86400000,
+		"maxExecutorCalls": 500,
+		"maxRepairPasses": 50,
+		"maxConcurrency": 2,
+		"maxContextChars": 120000
+	}
+}

package/docs/templates/evaluation/campaign-dogfood-v0.json

@@ -0,0 +1,24 @@
+{
+  "schemaVersion": 1,
+  "campaignId": "camp-dogfood-context-policy-ab",
+  "createdAt": "2026-07-19T00:00:00.000Z",
+  "status": "prepared",
+  "policyVersion": "promotion-policy-v1",
+  "incumbentCandidateId": "REPLACE_INCUMBENT_CANDIDATE_ID",
+  "challengerCandidateIds": ["REPLACE_CHALLENGER_CANDIDATE_ID"],
+  "corpusId": "corpus-dogfood-v0",
+  "corpusHash": "REPLACE_AFTER_CORPUS_REGISTER",
+  "privateVerifierId": "pv-dogfood-v0",
+  "privateVerifierManifestHash": "REPLACE_AFTER_PRIVATE_VERIFIER_REGISTER",
+  "budget": {
+    "schemaVersion": 1,
+    "mode": "hard",
+    "limits": {
+      "maxExecutorCalls": 500,
+      "maxWallTimeMs": 86400000,
+      "maxRepairPasses": 50,
+      "maxConcurrency": 2
+    }
+  },
+  "notes": "M4 dogfood: single-variable Context Policy A/B. Replace candidate/corpus/privateVerifier hashes after register. held_out cells are promotion_gate_only."
+}

package/docs/templates/evaluation/campaign-evidence-v1.json

@@ -0,0 +1,44 @@
+{
+  "schemaVersion": 1,
+  "campaignId": "campaign-dogfood-v0",
+  "challengerCandidateId": "cand-challenger-example",
+  "cells": [
+    {
+      "candidateId": "cand-baseline-example",
+      "taskRef": "synthetic/private-1",
+      "seed": 1,
+      "split": "private",
+      "verifyPassed": false
+    },
+    {
+      "candidateId": "cand-challenger-example",
+      "taskRef": "synthetic/private-1",
+      "seed": 1,
+      "split": "private",
+      "verifyPassed": true
+    },
+    {
+      "candidateId": "cand-baseline-example",
+      "taskRef": "synthetic/heldout-1",
+      "seed": 1,
+      "split": "held_out",
+      "verifyPassed": true
+    },
+    {
+      "candidateId": "cand-challenger-example",
+      "taskRef": "synthetic/heldout-1",
+      "seed": 1,
+      "split": "held_out",
+      "verifyPassed": true
+    }
+  ],
+  "diffs": [],
+  "verifyCommands": {
+    "incumbent": ["npm test", "npm run typecheck"],
+    "challenger": ["npm test", "npm run typecheck"]
+  },
+  "budgetStatus": {
+    "incumbent": "ok",
+    "challenger": "ok"
+  }
+}

package/docs/templates/evaluation/context-policy-baseline-v1.json

@@ -0,0 +1,17 @@
+{
+  "schemaVersion": 1,
+  "policyId": "baseline-v1",
+  "description": "Current DAG context assembly: shared upstream char budget, role skill defaults, learned patterns only for implementer.",
+  "knobs": {
+    "maxUpstreamCharsByRole": {
+      "default": 2000
+    },
+    "includeLearnedPatternsRoles": ["implementer"],
+    "perSkillMaxCharsByRole": {
+      "default": 3000
+    },
+    "totalMaxCharsByRole": {
+      "default": 12000
+    }
+  }
+}

package/docs/templates/evaluation/context-policy-role-specialized-v1.json

@@ -0,0 +1,28 @@
+{
+  "schemaVersion": 1,
+  "policyId": "role-specialized-v1",
+  "description": "Role-specialized upstream and skill-instruction budgets; skill names still resolve via baseline merge order.",
+  "knobs": {
+    "maxUpstreamCharsByRole": {
+      "default": 2000,
+      "scout": 1200,
+      "reviewer": 1200,
+      "implementer": 3000,
+      "verifier": 1600,
+      "closeout": 1600,
+      "planner": 2000,
+      "supervisor": 2000
+    },
+    "includeLearnedPatternsRoles": ["implementer", "closeout"],
+    "perSkillMaxCharsByRole": {
+      "default": 3000,
+      "scout": 2500,
+      "implementer": 3500
+    },
+    "totalMaxCharsByRole": {
+      "default": 12000,
+      "scout": 10000,
+      "implementer": 14000
+    }
+  }
+}

package/docs/templates/evaluation/corpus-dogfood-v0.manifest.json

@@ -0,0 +1,118 @@
+{
+  "schemaVersion": 1,
+  "corpusId": "corpus-dogfood-v0",
+  "createdAt": "2026-07-19T00:00:00.000Z",
+  "description": "W1.5 initial dogfood corpus contract: 15 heterogeneous tasks with public/private/held_out splits. Contract-only; not live campaign evidence.",
+  "tasks": [
+    {
+      "taskRef": "features/F-2026-001/login-flow",
+      "class": "fullstack",
+      "split": "public",
+      "seeds": [1, 2, 3],
+      "featureId": "F-2026-001",
+      "description": "Dogfood login Feature packet (public feedback allowed)."
+    },
+    {
+      "taskRef": "features/F-2026-002/profile-filter",
+      "class": "implement",
+      "split": "public",
+      "seeds": [1, 2],
+      "featureId": "F-2026-002",
+      "description": "Profile filter implementation path."
+    },
+    {
+      "taskRef": "features/F-2026-005/welcome-banner",
+      "class": "fullstack",
+      "split": "public",
+      "seeds": [1, 2, 3],
+      "featureId": "F-2026-005",
+      "description": "fullstack-v1 welcome sample (public)."
+    },
+    {
+      "taskRef": "synthetic/docs-agents-converge",
+      "class": "docs",
+      "split": "public",
+      "seeds": [1, 2],
+      "description": "Docs/AGENTS converge style task (synthetic contract slot)."
+    },
+    {
+      "taskRef": "synthetic/docs-changelog-summary",
+      "class": "docs",
+      "split": "public",
+      "seeds": [1, 2, 3],
+      "description": "Changelog-style documentation task."
+    },
+    {
+      "taskRef": "synthetic/frontend-welcome-render",
+      "class": "frontend",
+      "split": "public",
+      "seeds": [1, 2],
+      "description": "Frontend render/accessibility contract slot."
+    },
+    {
+      "taskRef": "features/F-2026-003/empty-profile-regression",
+      "class": "bugfix",
+      "split": "private",
+      "seeds": [1, 2, 3],
+      "featureId": "F-2026-003",
+      "description": "Empty profile id/displayName regression (private selection)."
+    },
+    {
+      "taskRef": "features/F-2026-004/versioned-bootstrap",
+      "class": "bootstrap",
+      "split": "private",
+      "seeds": [1, 2],
+      "featureId": "F-2026-004",
+      "description": "Versioned self-host bootstrap canary (private)."
+    },
+    {
+      "taskRef": "synthetic/backend-test-welcome-api",
+      "class": "backend-test",
+      "split": "private",
+      "seeds": [1, 2, 3],
+      "description": "Backend API test workflow contract slot."
+    },
+    {
+      "taskRef": "synthetic/refactor-outcome-tokens",
+      "class": "refactor",
+      "split": "private",
+      "seeds": [1, 2],
+      "description": "Bounded refactor around evidence token naming."
+    },
+    {
+      "taskRef": "synthetic/implement-ready-planner-gate",
+      "class": "implement",
+      "split": "private",
+      "seeds": [1, 2, 3],
+      "description": "Ready Planner artifact-gate implementation slot."
+    },
+    {
+      "taskRef": "synthetic/heldout-docs-handoff",
+      "class": "docs",
+      "split": "held_out",
+      "seeds": [1, 2],
+      "description": "Held-out docs/handoff generalization slot; never used for candidate generation."
+    },
+    {
+      "taskRef": "synthetic/heldout-bugfix-gate",
+      "class": "bugfix",
+      "split": "held_out",
+      "seeds": [1, 2, 3],
+      "description": "Held-out bugfix gate; promotion-only signal."
+    },
+    {
+      "taskRef": "synthetic/heldout-fullstack-smoke",
+      "class": "fullstack",
+      "split": "held_out",
+      "seeds": [1, 2],
+      "description": "Held-out fullstack smoke; not for public feedback."
+    },
+    {
+      "taskRef": "synthetic/heldout-bootstrap-slot",
+      "class": "bootstrap",
+      "split": "held_out",
+      "seeds": [1, 2, 3],
+      "description": "Held-out bootstrap identity/slot discipline."
+    }
+  ]
+}

package/docs/templates/evaluation/matrix-dag-dry-run-v1.json

@@ -0,0 +1,21 @@
+{
+  "version": 3,
+  "title": "eval-matrix-cell-dry-run",
+  "runtimeContract": {
+    "schemaVersion": 1,
+    "agentRuntime": "pi-only",
+    "repairWriterProtocol": "explicit-node-v1"
+  },
+  "tasks": [
+    {
+      "id": "matrix-cell-static",
+      "depends_on": [],
+      "complexity": "LOW",
+      "executor": "static",
+      "subtask_prompt": "Matrix cell placeholder; evaluation binding is injected by campaign matrix.",
+      "static": {
+        "resultMarkdown": "matrix cell dry-run placeholder"
+      }
+    }
+  ]
+}

package/docs/templates/evaluation/matrix-fixture-v1.json

@@ -0,0 +1,10 @@
+{
+  "schemaVersion": 1,
+  "defaultVerifyPassed": false,
+  "results": [
+    {
+      "split": "public",
+      "verifyPassed": true
+    }
+  ]
+}

package/docs/templates/evaluation/private-verifier-dogfood-v0.json

@@ -0,0 +1,16 @@
+{
+  "schemaVersion": 1,
+  "privateVerifierId": "pv-dogfood-v0",
+  "createdAt": "2026-07-19T00:00:00.000Z",
+  "description": "M4 dogfood private verifier. Controller-only; never enter candidate contentRefs/prompts/skill snapshots.",
+  "checks": [
+    {
+      "checkId": "node-version-ok",
+      "kind": "shell",
+      "command": ["node", "-e", "process.exit(process.versions.node ? 0 : 1)"],
+      "cwd": ".",
+      "timeoutMs": 15000,
+      "expectedExitCode": 0
+    }
+  ]
+}

package/docs/templates/product-line/AGENTS.md

@@ -4,5 +4,6 @@
 - Treat `constraints.allowed_paths` and `constraints.forbidden_paths` as machine-enforced write boundaries; review DAG writer `writeSet` before execution.
 - Advance only tasks whose dependencies are complete. Preserve failed run records; retries receive new worker run IDs.
 - A Ready task must have acceptance references, non-empty allowed/forbidden paths, and deterministic verification commands.
+- For `fullstack-v1` packets (`feature.yaml`), every TaskSpec should declare `execution.workflow`; required ACs need implementation/verification refs, `required_evidence`, and integration policy. Parallel writers must not share overlapping `allowed_paths`.
 - QA records an independent verdict and evidence. Do not write `status: success` closeout until `qa_verdict: pass` and non-empty `qa_evidence` exist.
 - Human gates remain human decisions. Record owner, time, reason, evidence, and follow-up without rewriting failed history.

package/docs/templates/product-line/README.md

@@ -7,3 +7,20 @@ agent-worker task validate-feature <feature-dir>
 ```
 
 The validator checks acceptance ID uniqueness, task references and dependencies, cycles, TaskSpec/path/verification completeness, and QA evidence before a successful closeout.
+
+## Profiles
+
+| Profile | How to declare | Behavior |
+|---|---|---|
+| `generic`（默认） | 省略 `feature.yaml`，或 `profile: generic` | 保持 legacy 校验；不强制 `execution.workflow` / 双覆盖 AC |
+| `fullstack-v1` | `feature.yaml` 中 `profile: fullstack-v1` 并声明 `scope` | 强制显式 workflow、required AC 双覆盖、frontend-test 依赖、writer writeSet 串行等结构门禁 |
+
+端到端全栈 dogfood 样板见仓库 `features/F-2026-005/`（欢迎语链路：契约 → BE/FE 实现 → backend-test / frontend-test → Final Verification）。
+
+## Minimal files
+
+- `feature.yaml`（可选；fullstack 必填）
+- `requirement.md` / `design.md` / `test-plan.md`
+- `acceptance.yaml`
+- `tasks/task-graph.yaml` + `tasks/*.yaml`
+- `links.md` / `closeout.yaml`（按交付阶段）

package/docs/templates/product-line/acceptance.yaml

@@ -12,3 +12,12 @@ acceptance:
       expected_task_refs:
         - BE-001
         - QA-001
+      # fullstack-v1 required AC fields (optional for generic packets):
+      # implementation_task_refs:
+      #   - BE-001
+      # verification_task_refs:
+      #   - BE-TEST-001
+      # required_evidence:
+      #   - backend-test-result-v1
+      #   - shell_verification
+      # integration: real-required

package/docs/templates/product-line/feature.yaml

@@ -0,0 +1,11 @@
+schema_version: 1
+feature_id: F-YYYY-NNN
+# generic = legacy/simple packet (default when this file is absent)
+# fullstack-v1 = enable structural gates for explicit workflows + dual-coverage ACs
+profile: generic
+# Uncomment and set true flags when profile is fullstack-v1:
+# scope:
+#   backend: true
+#   frontend: true
+#   backendVerification: true
+#   frontendVerification: true

package/docs/templates/product-line/task-graph.yaml

@@ -5,11 +5,19 @@ nodes:
     task: BE-001.yaml
     type: backend-feature
     depends_on: []
+    # Optional artifact-aware edges (Ready Planner envelope-level gate):
+    # produces:
+    #   - kind: backend-impl
+    #     schemaId: backend-impl-v1
   - id: QA-001
     task: QA-001.yaml
     type: qa-execute
     depends_on:
       - BE-001
+    # consumes:
+    #   - kind: backend-impl
+    #     schemaId: backend-impl-v1
+    #     producerTaskId: BE-001
 parallel_policy:
   max_parallel_tasks: 2
   disallow_same_file_parallel_writes: true

package/docs/templates/product-line/task.yaml

@@ -30,6 +30,10 @@ constraints:
     - <protected/path/**>
   hard_constraints:
     - Do not access production secrets
+# Explicit runtime workflow → taskKind. For fullstack-v1 packets this should be set
+# on every TaskSpec. Allowed: agent-dag | frontend-implementation | backend-test | frontend-test
+execution:
+  workflow: agent-dag
 verify:
   preset: standard
   mode: serial

package/harness.json

@@ -59,7 +59,7 @@
     "pi": {
       "description": "Pi planning, review, diagnosis, and bounded writing when DAG toolProfile=write",
       "LOW": "gpt-5.3-codex-spark",
-      "MED": "glm-5.2",
+      "MED": "grok-4.5",
       "HIGH": "gpt-5.6-sol"
     }
   }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tea-agent/loop-agent",
-  "version": "0.15.0",
+  "version": "0.16.0",
   "type": "module",
   "bin": {
     "loop-agent": "bin/loop-agent.js",