@tea-agent/loop-agent 0.4.0 → 0.5.0

package/docs/templates/agent-dag-decision-envelope.schema.json

@@ -1,213 +1,213 @@
-{
-  "$schema": "https://json-schema.org/draft/2020-12/schema",
-  "$id": "https://loop-agent.local/schemas/agent-dag-decision-envelope.schema.json",
-  "title": "Agent DAG Decision Envelope",
-  "description": "Structured output contract for advisory Agent DAG AI Secretary / Decision Gate nodes. M1 treats this as a documentation contract; M3 runtime parser extracts a ```DECISION_ENVELOPE_JSON fenced block and validates with zod plus semantic cross-field checks (recommendedOption must match options[].id; auto-approve must not set requiresHuman=true).",
-  "type": "object",
-  "additionalProperties": false,
-  "required": [
-    "schemaVersion",
-    "gateType",
-    "decisionScope",
-    "decision",
-    "confidence",
-    "riskLevel",
-    "requiresHuman",
-    "nextAction",
-    "policyVersion",
-    "policyChecks",
-    "rationale",
-    "evidence",
-    "blockingFindings",
-    "requiredRevisions",
-    "riskFlags",
-    "humanEscalation",
-    "audit"
-  ],
-  "properties": {
-    "schemaVersion": {
-      "type": "integer",
-      "const": 1
-    },
-    "gateType": {
-      "type": "string",
-      "enum": ["plan-gate", "side-effect-gate", "acceptance-gate", "closeout-gate"]
-    },
-    "decisionScope": {
-      "type": "string",
-      "enum": ["node", "rank", "dag-run", "task", "repo"]
-    },
-    "decision": {
-      "type": "string",
-      "enum": [
-        "auto-approve",
-        "approve-with-constraints",
-        "request-revision",
-        "run-more-verification",
-        "split-followup",
-        "reject",
-        "escalate-to-human",
-        "pause-wait-external"
-      ]
-    },
-    "confidence": {
-      "type": "number",
-      "minimum": 0,
-      "maximum": 1
-    },
-    "riskLevel": {
-      "type": "string",
-      "enum": ["low", "medium", "high", "critical"]
-    },
-    "requiresHuman": {
-      "type": "boolean"
-    },
-    "nextAction": {
-      "type": "string",
-      "enum": [
-        "continue",
-        "rerun-implement",
-        "rerun-verify",
-        "run-targeted-check",
-        "split-followup",
-        "pause-and-ask",
-        "abort"
-      ]
-    },
-    "policyVersion": {
-      "type": "string",
-      "minLength": 1
-    },
-    "policyChecks": {
-      "type": "object",
-      "additionalProperties": false,
-      "required": ["mustEscalateFlags", "evidenceComplete", "allowedAutoApprove"],
-      "properties": {
-        "mustEscalateFlags": {
-          "type": "array",
-          "items": { "type": "string", "minLength": 1 }
-        },
-        "evidenceComplete": {
-          "type": "boolean"
-        },
-        "allowedAutoApprove": {
-          "type": "boolean"
-        }
-      }
-    },
-    "rationale": {
-      "type": "array",
-      "minItems": 1,
-      "items": { "type": "string", "minLength": 1 }
-    },
-    "evidence": {
-      "type": "array",
-      "minItems": 1,
-      "items": {
-        "type": "object",
-        "additionalProperties": false,
-        "required": ["path", "kind", "status", "summary"],
-        "properties": {
-          "path": { "type": "string", "minLength": 1 },
-          "kind": { "type": "string", "minLength": 1 },
-          "status": {
-            "type": "string",
-            "enum": ["verified", "partial", "self-reported", "conflicting", "missing"]
-          },
-          "summary": { "type": "string", "minLength": 1 }
-        }
-      }
-    },
-    "blockingFindings": {
-      "type": "array",
-      "items": { "type": "string", "minLength": 1 }
-    },
-    "requiredRevisions": {
-      "type": "array",
-      "items": { "type": "string", "minLength": 1 }
-    },
-    "riskFlags": {
-      "type": "array",
-      "items": { "type": "string", "minLength": 1 }
-    },
-    "humanEscalation": {
-      "oneOf": [
-        { "type": "null" },
-        {
-          "type": "object",
-          "additionalProperties": false,
-          "required": ["question", "recommendedOption", "options"],
-          "properties": {
-            "question": { "type": "string", "minLength": 1 },
-            "recommendedOption": { "type": "string", "minLength": 1 },
-            "options": {
-              "type": "array",
-              "minItems": 1,
-              "items": {
-                "type": "object",
-                "additionalProperties": false,
-                "required": ["id", "label", "risk", "reason"],
-                "properties": {
-                  "id": { "type": "string", "minLength": 1 },
-                  "label": { "type": "string", "minLength": 1 },
-                  "risk": { "type": "string", "minLength": 1 },
-                  "reason": { "type": "string", "minLength": 1 }
-                }
-              }
-            }
-          }
-        }
-      ]
-    },
-    "audit": {
-      "type": "object",
-      "additionalProperties": true,
-      "required": ["runId", "nodeId", "model"],
-      "properties": {
-        "runId": { "type": "string", "minLength": 1 },
-        "nodeId": { "type": "string", "minLength": 1 },
-        "model": { "type": "string", "minLength": 1 },
-        "sourceHash": { "type": "string" }
-      }
-    }
-  },
-  "allOf": [
-    {
-      "if": {
-        "properties": { "requiresHuman": { "const": true } },
-        "required": ["requiresHuman"]
-      },
-      "then": {
-        "properties": {
-          "humanEscalation": { "type": "object" },
-          "nextAction": { "const": "pause-and-ask" }
-        },
-        "required": ["humanEscalation"]
-      }
-    },
-    {
-      "if": {
-        "properties": { "requiresHuman": { "const": false } },
-        "required": ["requiresHuman"]
-      },
-      "then": {
-        "properties": {
-          "humanEscalation": { "type": "null" }
-        }
-      }
-    },
-    {
-      "if": {
-        "properties": {
-          "decision": { "enum": ["escalate-to-human", "pause-wait-external"] }
-        },
-        "required": ["decision"]
-      },
-      "then": {
-        "properties": {
-          "requiresHuman": { "const": true }
-        }
-      }
-    }
-  ]
-}
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://loop-agent.local/schemas/agent-dag-decision-envelope.schema.json",
+  "title": "Agent DAG Decision Envelope",
+  "description": "Structured output contract for advisory Agent DAG AI Secretary / Decision Gate nodes. M1 treats this as a documentation contract; M3 runtime parser extracts a ```DECISION_ENVELOPE_JSON fenced block and validates with zod plus semantic cross-field checks (recommendedOption must match options[].id; auto-approve must not set requiresHuman=true).",
+  "type": "object",
+  "additionalProperties": false,
+  "required": [
+    "schemaVersion",
+    "gateType",
+    "decisionScope",
+    "decision",
+    "confidence",
+    "riskLevel",
+    "requiresHuman",
+    "nextAction",
+    "policyVersion",
+    "policyChecks",
+    "rationale",
+    "evidence",
+    "blockingFindings",
+    "requiredRevisions",
+    "riskFlags",
+    "humanEscalation",
+    "audit"
+  ],
+  "properties": {
+    "schemaVersion": {
+      "type": "integer",
+      "const": 1
+    },
+    "gateType": {
+      "type": "string",
+      "enum": ["plan-gate", "side-effect-gate", "acceptance-gate", "closeout-gate"]
+    },
+    "decisionScope": {
+      "type": "string",
+      "enum": ["node", "rank", "dag-run", "task", "repo"]
+    },
+    "decision": {
+      "type": "string",
+      "enum": [
+        "auto-approve",
+        "approve-with-constraints",
+        "request-revision",
+        "run-more-verification",
+        "split-followup",
+        "reject",
+        "escalate-to-human",
+        "pause-wait-external"
+      ]
+    },
+    "confidence": {
+      "type": "number",
+      "minimum": 0,
+      "maximum": 1
+    },
+    "riskLevel": {
+      "type": "string",
+      "enum": ["low", "medium", "high", "critical"]
+    },
+    "requiresHuman": {
+      "type": "boolean"
+    },
+    "nextAction": {
+      "type": "string",
+      "enum": [
+        "continue",
+        "rerun-implement",
+        "rerun-verify",
+        "run-targeted-check",
+        "split-followup",
+        "pause-and-ask",
+        "abort"
+      ]
+    },
+    "policyVersion": {
+      "type": "string",
+      "minLength": 1
+    },
+    "policyChecks": {
+      "type": "object",
+      "additionalProperties": false,
+      "required": ["mustEscalateFlags", "evidenceComplete", "allowedAutoApprove"],
+      "properties": {
+        "mustEscalateFlags": {
+          "type": "array",
+          "items": { "type": "string", "minLength": 1 }
+        },
+        "evidenceComplete": {
+          "type": "boolean"
+        },
+        "allowedAutoApprove": {
+          "type": "boolean"
+        }
+      }
+    },
+    "rationale": {
+      "type": "array",
+      "minItems": 1,
+      "items": { "type": "string", "minLength": 1 }
+    },
+    "evidence": {
+      "type": "array",
+      "minItems": 1,
+      "items": {
+        "type": "object",
+        "additionalProperties": false,
+        "required": ["path", "kind", "status", "summary"],
+        "properties": {
+          "path": { "type": "string", "minLength": 1 },
+          "kind": { "type": "string", "minLength": 1 },
+          "status": {
+            "type": "string",
+            "enum": ["verified", "partial", "self-reported", "conflicting", "missing"]
+          },
+          "summary": { "type": "string", "minLength": 1 }
+        }
+      }
+    },
+    "blockingFindings": {
+      "type": "array",
+      "items": { "type": "string", "minLength": 1 }
+    },
+    "requiredRevisions": {
+      "type": "array",
+      "items": { "type": "string", "minLength": 1 }
+    },
+    "riskFlags": {
+      "type": "array",
+      "items": { "type": "string", "minLength": 1 }
+    },
+    "humanEscalation": {
+      "oneOf": [
+        { "type": "null" },
+        {
+          "type": "object",
+          "additionalProperties": false,
+          "required": ["question", "recommendedOption", "options"],
+          "properties": {
+            "question": { "type": "string", "minLength": 1 },
+            "recommendedOption": { "type": "string", "minLength": 1 },
+            "options": {
+              "type": "array",
+              "minItems": 1,
+              "items": {
+                "type": "object",
+                "additionalProperties": false,
+                "required": ["id", "label", "risk", "reason"],
+                "properties": {
+                  "id": { "type": "string", "minLength": 1 },
+                  "label": { "type": "string", "minLength": 1 },
+                  "risk": { "type": "string", "minLength": 1 },
+                  "reason": { "type": "string", "minLength": 1 }
+                }
+              }
+            }
+          }
+        }
+      ]
+    },
+    "audit": {
+      "type": "object",
+      "additionalProperties": true,
+      "required": ["runId", "nodeId", "model"],
+      "properties": {
+        "runId": { "type": "string", "minLength": 1 },
+        "nodeId": { "type": "string", "minLength": 1 },
+        "model": { "type": "string", "minLength": 1 },
+        "sourceHash": { "type": "string" }
+      }
+    }
+  },
+  "allOf": [
+    {
+      "if": {
+        "properties": { "requiresHuman": { "const": true } },
+        "required": ["requiresHuman"]
+      },
+      "then": {
+        "properties": {
+          "humanEscalation": { "type": "object" },
+          "nextAction": { "const": "pause-and-ask" }
+        },
+        "required": ["humanEscalation"]
+      }
+    },
+    {
+      "if": {
+        "properties": { "requiresHuman": { "const": false } },
+        "required": ["requiresHuman"]
+      },
+      "then": {
+        "properties": {
+          "humanEscalation": { "type": "null" }
+        }
+      }
+    },
+    {
+      "if": {
+        "properties": {
+          "decision": { "enum": ["escalate-to-human", "pause-wait-external"] }
+        },
+        "required": ["decision"]
+      },
+      "then": {
+        "properties": {
+          "requiresHuman": { "const": true }
+        }
+      }
+    }
+  ]
+}

package/docs/templates/agent-dag-decision-gate-dogfood-report.md

@@ -1,117 +1,117 @@
-# Agent DAG Decision Gate Dogfood Report Template
-
-> Copy to `docs/reports/YYYY-MM-DD-agent-dag-decision-gate-dogfood.md` after each dogfood run.
-
-## Run Metadata
-
-| Field | Value |
-|---|---|
-| Date | YYYY-MM-DD |
-| Topic |  |
-| DAG input | `<temp-dir>/<topic>-decision-gate-dag.json` |
-| Run ID |  |
-| Run facts | `.harness/dag-runs/completed/<run-id>/` |
-| Gate type | `acceptance-gate` |
-| Decision node | `decision-pi` |
-| Model | `gpt-5.5` via `executorModels.pi.HIGH` |
-
-## Work Type
-
-Choose one:
-
-- [ ] Low-risk docs/template change — expected `auto-approve` or `approve-with-constraints`
-- [ ] Runtime/loop-agent change — expected `request-revision`, `run-more-verification`, or `auto-approve`
-- [ ] Contract/security/high-risk simulation — expected `escalate-to-human`
-
-## Deterministic Evidence
-
-| Evidence | Status | Notes |
-|---|---|---|
-| `verify-shell/result.summary.md` | verified / partial / missing |  |
-| `state.json` | verified / partial / missing |  |
-| git diff / changed file list | verified / partial / missing |  |
-| progress/report/artifacts | verified / partial / missing |  |
-
-## Decision Envelope Summary
-
-Paste the **exact** ` ```DECISION_ENVELOPE_JSON ` fenced block from `decision-pi/result.summary.md` (info string must be `DECISION_ENVELOPE_JSON`, not `json`). Also record `decision`, `requiresHuman`, `nextAction`, and verified evidence count.
-
-```DECISION_ENVELOPE_JSON
-{
-  "schemaVersion": 1,
-  "gateType": "acceptance-gate",
-  "decisionScope": "dag-run",
-  "decision": "auto-approve",
-  "confidence": 0.88,
-  "riskLevel": "low",
-  "requiresHuman": false,
-  "nextAction": "continue",
-  "policyVersion": "agent-dag-decision-gate-v1",
-  "policyChecks": {
-    "mustEscalateFlags": [],
-    "evidenceComplete": true,
-    "allowedAutoApprove": true
-  },
-  "rationale": ["..."],
-  "evidence": [
-    {
-      "path": ".harness/dag-runs/completed/<run-id>/verify-shell/result.summary.md",
-      "kind": "shell-output",
-      "status": "verified",
-      "summary": "verification commands passed"
-    }
-  ],
-  "blockingFindings": [],
-  "requiredRevisions": [],
-  "riskFlags": [],
-  "humanEscalation": null,
-  "audit": {
-    "runId": "<run-id>",
-    "nodeId": "decision-pi",
-    "model": "gpt-5.5"
-  }
-}
-```
-
-## Quality Metrics
-
-| Metric | Value | Notes |
-|---|---:|---|
-| `decisionLatencyMs` |  | From node duration |
-| `tokenCostEstimate` |  | If available |
-| `humanInterruptCount` |  | Should be 0 for low-risk auto decisions |
-| `falseEscalation` | yes / no | Escalated when policy allowed auto decision |
-| `missedEscalation` | yes / no | Auto-approved when policy required human escalation |
-| `revisionCaughtBeforeHuman` | yes / no | Gate requested revision before human involvement |
-| `verifyPassAfterGate` | yes / no | Later verification passed after gate action |
-| `evidenceCompleteness` | complete / partial / missing | Based on verified evidence count |
-
-## Evaluation
-
-### What the gate got right
-
-- 
-
-### What the gate got wrong or over/under-weighted
-
-- 
-
-### Prompt / policy adjustments needed
-
-- 
-
-## Recommendation
-
-Choose one:
-
-- [ ] `proceed` — enough evidence to move toward parser/runtime work
-- [ ] `needs-more-dogfood` — continue M1/M2 advisory runs
-- [ ] `defer` — gate quality not yet good enough
-
-Rationale:
-
-- 
-
-## Follow-ups
-
-- 
+# Agent DAG Decision Gate Dogfood Report Template
+
+> Copy to `docs/reports/YYYY-MM-DD-agent-dag-decision-gate-dogfood.md` after each dogfood run.
+
+## Run Metadata
+
+| Field | Value |
+|---|---|
+| Date | YYYY-MM-DD |
+| Topic |  |
+| DAG input | `<temp-dir>/<topic>-decision-gate-dag.json` |
+| Run ID |  |
+| Run facts | `.harness/dag-runs/completed/<run-id>/` |
+| Gate type | `acceptance-gate` |
+| Decision node | `decision-pi` |
+| Model | `gpt-5.5` via `executorModels.pi.HIGH` |
+
+## Work Type
+
+Choose one:
+
+- [ ] Low-risk docs/template change — expected `auto-approve` or `approve-with-constraints`
+- [ ] Runtime/loop-agent change — expected `request-revision`, `run-more-verification`, or `auto-approve`
+- [ ] Contract/security/high-risk simulation — expected `escalate-to-human`
+
+## Deterministic Evidence
+
+| Evidence | Status | Notes |
+|---|---|---|
+| `verify-shell/result.summary.md` | verified / partial / missing |  |
+| `state.json` | verified / partial / missing |  |
+| git diff / changed file list | verified / partial / missing |  |
+| progress/report/artifacts | verified / partial / missing |  |
+
+## Decision Envelope Summary
+
+Paste the **exact** ` ```DECISION_ENVELOPE_JSON ` fenced block from `decision-pi/result.summary.md` (info string must be `DECISION_ENVELOPE_JSON`, not `json`). Also record `decision`, `requiresHuman`, `nextAction`, and verified evidence count.
+
+```DECISION_ENVELOPE_JSON
+{
+  "schemaVersion": 1,
+  "gateType": "acceptance-gate",
+  "decisionScope": "dag-run",
+  "decision": "auto-approve",
+  "confidence": 0.88,
+  "riskLevel": "low",
+  "requiresHuman": false,
+  "nextAction": "continue",
+  "policyVersion": "agent-dag-decision-gate-v1",
+  "policyChecks": {
+    "mustEscalateFlags": [],
+    "evidenceComplete": true,
+    "allowedAutoApprove": true
+  },
+  "rationale": ["..."],
+  "evidence": [
+    {
+      "path": ".harness/dag-runs/completed/<run-id>/verify-shell/result.summary.md",
+      "kind": "shell-output",
+      "status": "verified",
+      "summary": "verification commands passed"
+    }
+  ],
+  "blockingFindings": [],
+  "requiredRevisions": [],
+  "riskFlags": [],
+  "humanEscalation": null,
+  "audit": {
+    "runId": "<run-id>",
+    "nodeId": "decision-pi",
+    "model": "gpt-5.5"
+  }
+}
+```
+
+## Quality Metrics
+
+| Metric | Value | Notes |
+|---|---:|---|
+| `decisionLatencyMs` |  | From node duration |
+| `tokenCostEstimate` |  | If available |
+| `humanInterruptCount` |  | Should be 0 for low-risk auto decisions |
+| `falseEscalation` | yes / no | Escalated when policy allowed auto decision |
+| `missedEscalation` | yes / no | Auto-approved when policy required human escalation |
+| `revisionCaughtBeforeHuman` | yes / no | Gate requested revision before human involvement |
+| `verifyPassAfterGate` | yes / no | Later verification passed after gate action |
+| `evidenceCompleteness` | complete / partial / missing | Based on verified evidence count |
+
+## Evaluation
+
+### What the gate got right
+
+- 
+
+### What the gate got wrong or over/under-weighted
+
+- 
+
+### Prompt / policy adjustments needed
+
+- 
+
+## Recommendation
+
+Choose one:
+
+- [ ] `proceed` — enough evidence to move toward parser/runtime work
+- [ ] `needs-more-dogfood` — continue M1/M2 advisory runs
+- [ ] `defer` — gate quality not yet good enough
+
+Rationale:
+
+- 
+
+## Follow-ups
+
+-