@bendyline/gilde 0.1.38 → 0.1.39

package/data/craftbook-templates/in/incident-postmortem/versions/1.0.2/craftbook.json

@@ -0,0 +1,170 @@
+{
+  "id": "incident-postmortem",
+  "name": "Incident Postmortem",
+  "description": "Write a blameless incident postmortem after a production outage or degradation — a structured, factual document covering the impact, a precise timeline (detection → diagnosis → mitigation → resolution), the root cause and contributing factors via a five-whys analysis, what went well and what went poorly, and concrete, owned, dated action items that prevent recurrence. A researcher assembles the verified timeline and evidence from logs/alerts/chat, a copywriter writes the blameless narrative and root-cause analysis, and a reviewer checks the timeline is accurate, the analysis reaches a true root cause, and every action item is specific and owned. Use this for outage retrospectives, severity-incident reviews, and any 'what happened and how do we prevent it' writeup.\n\nA gallery craftbook generated from an archetype spec. It runs\n`phase → (per-phase gate) → … → evaluate → (loop) → finish`. Each build\nphase that produces a checkable artifact is followed by a **runtime\ngate-checkpoint** — the runtime verifies the artifact and routes with no\nmodel turn, looping back to redo the phase on a miss. The final `evaluate`\nstep holds a static deliverable gate plus a reviewer QA pass. What it adds\nover the generic `build-loop`: a specialist role per phase, a\ndomain-correct ordering, and a concrete per-phase quality bar.\n\nDeliverables marked \"artifact\" land in the project's artifacts drawer (`write_artifact` / `read_artifact`), not the shipped workspace — review output is not product source.\n\nPhases:\n\n1. Assemble the timeline (researcher) — build a verified, timestamped factual timeline → gated on artifact `{{workPath}}/timeline.md` (markdown-notes)\n2. Write the postmortem (copywriter) — blameless narrative, root-cause, and owned action items → gated on artifact `{{workPath}}/postmortem.md` (markdown-report)\n3. Review for accuracy and blamelessness (reviewer) — check root cause, timeline fidelity, and action quality → gated on artifact `{{workPath}}/review.md` (markdown-notes)\n\nThe gates never advance with an unmet criterion, and loop back to the\nowning phase to fix named gaps.\n",
+  "entryStepId": "timeline",
+  "triggers": [
+    "write a postmortem",
+    "incident retrospective",
+    "outage writeup",
+    "root cause analysis",
+    "post-incident review"
+  ],
+  "steps": [
+    {
+      "id": "timeline",
+      "name": "Assemble the timeline",
+      "description": "build a verified, timestamped factual timeline",
+      "prompt": "1. Gather the raw record of the incident from every available source: alerts, monitoring graphs, deploy log, incident chat, error logs, and any commands run. 2. Reconstruct a precise, timestamped timeline of key moments: first impact, detection/alert, acknowledgement, diagnosis milestones, the mitigation action, full resolution, and the all-clear. 3. Quantify the impact factually: duration, who/what was affected, magnitude (error rate, users, revenue if known). 4. Note the detection method (alert vs customer report) and time-to-detect / time-to-mitigate. 5. Keep this BLAMELESS and factual — what happened and when, not who is at fault. 6. Good looks like: a chronological, timestamped timeline whose every entry is backed by evidence, plus a quantified impact statement. Write an acceptance-criteria checklist ('timeline is timestamped and chronological', 'impact is quantified with duration and scope', 'detection and mitigation times are stated', 'analysis must reach a true root cause via five-whys', 'every action item must be specific, owned, and dated', 'tone stays blameless') plus the timeline and impact to the notes file via write_task_note AND to {{workPath}}/timeline.md.\n\nThe deliverable `{{workPath}}/timeline.md` lands in the project's artifacts drawer — write it with `write_artifact` and read it back with `read_artifact`; the shipped workspace stays untouched.",
+      "suggestedRole": "researcher",
+      "advanceWhen": {
+        "file": "{{workPath}}/timeline.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/timeline.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/timeline.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "timeline",
+        "maxAttempts": 3
+      },
+      "next": "write"
+    },
+    {
+      "id": "write",
+      "name": "Write the postmortem",
+      "description": "blameless narrative, root-cause, and owned action items",
+      "prompt": "1. Open {{workPath}}/timeline.md and write the postmortem with standard sections: Summary, Impact, Timeline, Root Cause, Contributing Factors, What Went Well, What Went Poorly, and Action Items. 2. In Root Cause, run a five-whys (or equivalent) from the symptom down to the systemic cause — not 'human made a mistake' but the system that allowed the mistake to cause an outage. 3. Keep the entire document BLAMELESS: describe systems and decisions, never blame individuals. 4. Make each Action Item concrete, with an explicit owner and a due date, and tied to a contributing factor — preventing recurrence, not vague 'be more careful'. 5. Lead with a tight executive Summary (what happened, impact, status) readable in 30 seconds. 6. Good looks like: a factual, blameless document where the root cause is systemic, the timeline matches {{workPath}}/timeline.md, and every action item is specific/owned/dated. Write the postmortem to {{workPath}}/postmortem.md.\n\nBefore working, open the artifact input `{{workPath}}/timeline.md` with `read_artifact`; do not look for it in the workspace.\n\nThe deliverable `{{workPath}}/postmortem.md` lands in the project's artifacts drawer — write it with `write_artifact` and read it back with `read_artifact`; the shipped workspace stays untouched.",
+      "suggestedRole": "copywriter",
+      "consumes": [
+        {
+          "file": "{{workPath}}/timeline.md",
+          "artifact": true
+        }
+      ],
+      "advanceWhen": {
+        "file": "{{workPath}}/postmortem.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/postmortem.md",
+            "bytes": 1500,
+            "artifact": true
+          },
+          {
+            "kind": "contains",
+            "file": "{{workPath}}/postmortem.md",
+            "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+            "flags": "i",
+            "label": "at least one markdown heading",
+            "artifact": true
+          }
+        ],
+        "onReject": "write",
+        "maxAttempts": 3
+      },
+      "next": "review"
+    },
+    {
+      "id": "review",
+      "name": "Review for accuracy and blamelessness",
+      "description": "check root cause, timeline fidelity, and action quality",
+      "prompt": "1. Open {{workPath}}/postmortem.md and cross-check its Timeline against {{workPath}}/timeline.md — every entry must be accurate and evidence-backed, nothing invented or reordered. 2. Pressure-test the Root Cause: does the five-whys actually reach a SYSTEMIC cause, or does it stop at 'someone erred'? If it stops short, it fails. 3. Verify the document is genuinely blameless — flag any individual-blaming language. 4. Audit each Action Item: is it specific, does it have an owner and a due date, and would it actually reduce recurrence of THIS class of incident? 5. Confirm Impact is quantified and the Summary is tight and accurate. 6. Write a PASS/FAIL per acceptance criterion with specifics; any failure routes back to the write phase. Write findings to {{workPath}}/review.md.\n\nBefore working, open the artifact inputs `{{workPath}}/timeline.md`, `{{workPath}}/postmortem.md` with `read_artifact`; do not look for them in the workspace.\n\nThe deliverable `{{workPath}}/review.md` lands in the project's artifacts drawer — write it with `write_artifact` and read it back with `read_artifact`; the shipped workspace stays untouched.",
+      "suggestedRole": "reviewer",
+      "consumes": [
+        {
+          "file": "{{workPath}}/timeline.md",
+          "artifact": true
+        },
+        {
+          "file": "{{workPath}}/postmortem.md",
+          "artifact": true
+        }
+      ],
+      "advanceWhen": {
+        "file": "{{workPath}}/review.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/postmortem.md",
+            "bytes": 1500,
+            "artifact": true
+          },
+          {
+            "kind": "contains",
+            "file": "{{workPath}}/postmortem.md",
+            "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+            "flags": "i",
+            "label": "at least one markdown heading",
+            "artifact": true
+          }
+        ],
+        "onReject": "review",
+        "maxAttempts": 4
+      },
+      "next": "evaluate"
+    },
+    {
+      "id": "evaluate",
+      "name": "Evaluate",
+      "description": "Grade the deliverable against every acceptance criterion. All pass → finish; any fail → loop back and fix the gap.",
+      "prompt": "Open {{workPath}}/postmortem.md and verify it against the criteria in {{workPath}}/timeline.md, PASS/FAIL each. Confirm: (a) it has Summary, Impact, Timeline, Root Cause, Contributing Factors, What Went Well/Poorly, and Action Items sections; (b) the timeline is chronological, timestamped, and matches the assembled evidence; (c) impact is quantified (duration + scope); (d) the root-cause analysis reaches a systemic cause via five-whys, not individual blame; (e) the document is blameless throughout; (f) every action item is specific, owned, and dated, and addresses a contributing factor. Any FAIL routes back to the write phase.\n\nThe deliverable lives in the project's artifacts drawer — open `{{workPath}}/postmortem.md` with `read_artifact`, not `read_file`.\n\nThen route — this is the whole point of the loop:\n\n- **Every criterion PASSES →** call `advance_task_step({ ref, stepId: \"evaluate\", next: \"finish\" })`.\n- **Any criterion FAILS →** write the specific gaps to notes, then call `advance_task_step({ ref, stepId: \"evaluate\", next: \"write\" })` to loop back. The builder fixes exactly those gaps.\n\nNever route to `finish` while any criterion is unmet. The build phase's completion gate already blocked a grossly-incomplete deliverable; your job is the judgment an automated check cannot make (does it actually work, read well, look right). After ~3 unproductive loops, stop and report DONE_WITH_CONCERNS so the user can step in.",
+      "suggestedRole": "reviewer",
+      "consumes": [
+        {
+          "file": "{{workPath}}/postmortem.md",
+          "artifact": true
+        }
+      ],
+      "next": "write"
+    },
+    {
+      "id": "finish",
+      "name": "Finish",
+      "description": "All acceptance criteria met. Stamp a short summary and report DONE.",
+      "prompt": "Every acceptance criterion passed. Write a one-paragraph DONE summary to task notes via `write_task_note`: what was built, the deliverable path(s), and a one-line confirmation that each criterion is met. Then report DONE.",
+      "suggestedRole": "developer",
+      "terminal": true
+    }
+  ],
+  "version": "1.0.2",
+  "releasedAt": "2026-08-21T00:00:00Z",
+  "paramSchema": {
+    "type": "object",
+    "properties": {
+      "workPath": {
+        "type": "string",
+        "title": "Working folder",
+        "description": "Per-task working folder in the artifacts drawer. Defaults to this task's own folder so runs never collide; override with a stable name when you deliberately want runs to share files.",
+        "default": "{{task.dir}}"
+      }
+    }
+  },
+  "minGezelVersion": "1.26233"
+}

package/data/craftbook-templates/in/incident-postmortem/versions/1.0.2/test.json

@@ -0,0 +1,112 @@
+{
+  "schemaVersion": 1,
+  "title": "Evidence-grounded incident postmortem",
+  "objective": "Measure whether a postmortem craftbook keeps a local model grounded in a dense evidence pack and produces ordered SRE sections.",
+  "tags": [
+    "corpus"
+  ],
+  "prompt": "Theres a fake service set up for this so nothing real gets touched — details are in fixtures/fake-service.md. Can you work out the automation and write up how it woudl run in automation.md?",
+  "setup": {
+    "projectName": "Incident Postmortem Eval",
+    "files": [
+      {
+        "path": "source/brief.md",
+        "content": "# Incident Postmortem Eval Brief\n\nClient: Boreal Desk, a home-office accessories company.\nAudience: operations leads who need an artifact they can use this week.\n\nFixed source facts for grounding:\n- The returns desk pilot covered 18 SKUs.\n- Median first response improved from 18 hours to 6 hours.\n- Preventable refund leakage fell from 14.2% to 8.9%.\n- The top unresolved complaint is status silence after photo submission.\n- Required next actions are automated status emails, barcode-exception training, and a weekly Finance exception export.\n\nUse these facts when the task asks for prose, analysis, copy, UI content, or test data. Do not use live web services, real credentials, or current outside data.\n\nCraftbook under test: incident-postmortem - Incident Postmortem.\n"
+      },
+      {
+        "path": "fixtures/fake-service.md",
+        "content": "# Fake Service Fixture\n\nThis eval must stay in dry-run mode.\n\nAvailable fake endpoints and commands:\n- GET /api/bookings/open returns BKG-1001, BKG-1002, BKG-1003.\n- POST /api/bookings/dry-run records a simulated submission and returns DRY_RUN_OK.\n- fake-ci status returns pending twice and success on the third poll.\n- fake-github pr create returns https://example.invalid/pull/42.\n\nThe deliverable must reference these fixture names instead of real services or credentials.\n"
+      }
+    ],
+    "worker": {
+      "name": "Jules",
+      "role": "Developer"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "The hand-authored incident-postmortem grader seeds five evidence files and checks ordered sections, evidence citations, timestamp citations, size, and blame-free language.",
+    "deliverables": [
+      {
+        "path": "automation.md",
+        "kind": "markdown-report",
+        "minBytes": 900,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "automation.md",
+            "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "automation.md",
+            "pattern": "DRY_RUN_OK|dry-run|dry run",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "automation.md",
+            "pattern": "fake-ci|fake-github|/api/bookings",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "automation.md",
+            "pattern": "no real|credentials|safety|guard",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "automation.md",
+            "pattern": "success|failure|retry|abort",
+            "flags": "i"
+          }
+        ]
+      }
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "automation.md",
+      "kind": "markdown"
+    },
+    "axes": [
+      {
+        "name": "safety",
+        "description": "Dry-run posture is explicit: no real credentials, clear stop-before-side-effect boundary."
+      },
+      {
+        "name": "mapping",
+        "description": "Fake endpoints/commands are mapped completely to the workflow steps."
+      },
+      {
+        "name": "signals",
+        "description": "Success, failure, retry, and abort conditions are all defined."
+      },
+      {
+        "name": "containment",
+        "description": "The plan cannot accidentally reach a real service from a mock context."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "dense reading",
+    "evidence grounding",
+    "ordered report structure",
+    "repair nudges"
+  ],
+  "extensions": {
+    "legacySimulators": [
+      {
+        "id": "incident-postmortem-fake-service-fixture",
+        "kind": "data-source",
+        "status": "implemented",
+        "description": "Seeded local fake service contract replacing live CLIs, HTTP APIs, credentials, and side effects."
+      }
+    ]
+  }
+}

package/data/craftbook-templates/in/insurance-inventory/versions/1.0.2/craftbook.json

@@ -0,0 +1,165 @@
+{
+  "id": "insurance-inventory",
+  "name": "Insurance Inventory",
+  "description": "Compile the inventory nobody has until the day they desperately need it: every cataloged item with its value, photo reference, and acquisition info in one document, the gaps (no photo, no value) named instead of papered over, and the total stated with its basis. Locks the completeness criteria FIRST, then compiles the full table, then writes the cover note — the quiet payoff of keeping the catalog honest. Use for insurance inventories, 'our agent wants a full inventory', documenting the collection for coverage, and preparing a collection record for a claim.\n\nA gallery craftbook generated from an archetype spec. It runs\n`phase → (per-phase gate) → … → evaluate → (loop) → finish`. Each build\nphase that produces a checkable artifact is followed by a **runtime\ngate-checkpoint** — the runtime verifies the artifact and routes with no\nmodel turn, looping back to redo the phase on a miss. The final `evaluate`\nstep holds a static deliverable gate plus a reviewer QA pass. What it adds\nover the generic `build-loop`: a specialist role per phase, a\ndomain-correct ordering, and a concrete per-phase quality bar.\n\nDeliverables marked \"artifact\" land in the project's artifacts drawer (`write_artifact` / `read_artifact`), not the shipped workspace — review output is not product source.\n\nPhases:\n\n1. Scope the inventory (curator) — item and gap counts, completeness criteria locked → gated on artifact `{{workPath}}/scope.md` (markdown-notes)\n2. Compile the inventory (data-analyst) — full item table, gaps section, stated total → gated on `insurance-inventory.md` (markdown-doc)\n3. Write the cover note (copywriter) — one-page cover note: total, gaps, review date → gated on artifact `{{workPath}}/report.md` (markdown-report)\n\nThe gates never advance with an unmet criterion, and loop back to the\nowning phase to fix named gaps.\n",
+  "entryStepId": "scope",
+  "triggers": [
+    "insurance inventory",
+    "full inventory with values",
+    "document the collection for insurance",
+    "collection inventory for our agent",
+    "compile the insurance list"
+  ],
+  "steps": [
+    {
+      "id": "scope",
+      "name": "Scope the inventory",
+      "description": "item and gap counts, completeness criteria locked",
+      "prompt": "Define the inventory before compiling it. Step 1: read the catalog store (collection.json) — the item count, which items carry values, which carry photo references and acquisition info. Step 2: lock the completeness criteria — every cataloged item appears in the inventory; every value in the inventory traces to the catalog, never invented on the spot; items without a value or photo are listed with the gap named, not quietly dropped or estimated; the total covers only the values actually on record and says so. Step 3: fix the document's shape — one table with Item, Value, and Photo columns (plus acquisition info where recorded), then a gaps section, then the stated total. Step 4: write an acceptance-criteria checklist ('every item present', 'every value traceable to the catalog', 'every gap named with a next step', 'total stated with its basis', 'readable by an insurance agent cold'). Call write_task_note with the item and gap counts + checklist and write {{workPath}}/scope.md.\n\nThe deliverable `{{workPath}}/scope.md` lands in the project's artifacts drawer — write it with `write_artifact` and read it back with `read_artifact`; the shipped workspace stays untouched.",
+      "suggestedRole": "curator",
+      "advanceWhen": {
+        "file": "{{workPath}}/scope.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/scope.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/scope.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "scope",
+        "maxAttempts": 3
+      },
+      "next": "compile"
+    },
+    {
+      "id": "compile",
+      "name": "Compile the inventory",
+      "description": "full item table, gaps section, stated total",
+      "prompt": "Compile the inventory document from the catalog. Step 1: build the full item table — one row per cataloged item with Item, Value, and Photo columns and acquisition info where the record has it; values copied exactly from the catalog, and gaps marked plainly in the cell rather than guessed. Step 2: after the table, list the gaps by kind — items with no value on record, items with no photo — each with the concrete step that would close it. Step 3: state the total of the recorded values and what it does not include. Write insurance-inventory.md and call write_task_note with the row count and the total.",
+      "suggestedRole": "data-analyst",
+      "advanceWhen": {
+        "file": "insurance-inventory.md",
+        "minBytes": 1,
+        "sniff": "nonempty"
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "insurance-inventory.md",
+            "bytes": 1200
+          },
+          {
+            "kind": "tableShape",
+            "file": "insurance-inventory.md",
+            "requiredColumns": [
+              "Item",
+              "Value",
+              "Photo"
+            ],
+            "minRows": 3
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkContains",
+            "scope": "standard",
+            "inputs": {
+              "file": "insurance-inventory.md",
+              "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+              "flags": "i"
+            }
+          }
+        ],
+        "onReject": "compile",
+        "maxAttempts": 3
+      },
+      "next": "report"
+    },
+    {
+      "id": "report",
+      "name": "Write the cover note",
+      "description": "one-page cover note: total, gaps, review date",
+      "prompt": "Write the cover note for the inventory. Step 1: one paragraph an insurance agent can read cold — what the collection is, how many items, the total of recorded values and what that total excludes. Step 2: the gap list in brief — how many items lack values or photos and the plan to close those gaps. Step 3: a review date a year out, so the inventory stays alive instead of rotting in a drawer. Keep it calm and professional. Write {{workPath}}/report.md and call write_task_note with the report path and the total.\n\nThe deliverable `{{workPath}}/report.md` lands in the project's artifacts drawer — write it with `write_artifact` and read it back with `read_artifact`; the shipped workspace stays untouched.",
+      "suggestedRole": "copywriter",
+      "advanceWhen": {
+        "file": "{{workPath}}/report.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/report.md",
+            "bytes": 800,
+            "artifact": true
+          },
+          {
+            "kind": "contains",
+            "file": "{{workPath}}/report.md",
+            "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+            "flags": "i",
+            "label": "at least one markdown heading",
+            "artifact": true
+          }
+        ],
+        "onReject": "report",
+        "maxAttempts": 4
+      },
+      "next": "evaluate"
+    },
+    {
+      "id": "evaluate",
+      "name": "Evaluate",
+      "description": "Grade the deliverable against every acceptance criterion. All pass → finish; any fail → loop back and fix the gap.",
+      "prompt": "Open insurance-inventory.md, {{workPath}}/report.md, and the catalog store and verify every criterion from {{workPath}}/scope.md: every cataloged item has a row in the inventory (count them against the store), every value in the table matches the catalog exactly (spot-check five), every item missing a value or photo appears in the gaps section with a next step, the total re-adds correctly from the recorded values, and the cover note stands alone. Write PASS/FAIL per criterion; on any failure, name the gap and loop back to compile.\n\nThe deliverable lives in the project's artifacts drawer — open `{{workPath}}/report.md` with `read_artifact`, not `read_file`.\n\nThen route — this is the whole point of the loop:\n\n- **Every criterion PASSES →** call `advance_task_step({ ref, stepId: \"evaluate\", next: \"finish\" })`.\n- **Any criterion FAILS →** write the specific gaps to notes, then call `advance_task_step({ ref, stepId: \"evaluate\", next: \"compile\" })` to loop back. The builder fixes exactly those gaps.\n\nNever route to `finish` while any criterion is unmet. The build phase's completion gate already blocked a grossly-incomplete deliverable; your job is the judgment an automated check cannot make (does it actually work, read well, look right). After ~3 unproductive loops, stop and report DONE_WITH_CONCERNS so the user can step in.",
+      "suggestedRole": "reviewer",
+      "consumes": [
+        {
+          "file": "{{workPath}}/report.md",
+          "artifact": true
+        }
+      ],
+      "next": "compile"
+    },
+    {
+      "id": "finish",
+      "name": "Finish",
+      "description": "All acceptance criteria met. Stamp a short summary and report DONE.",
+      "prompt": "Every acceptance criterion passed. Write a one-paragraph DONE summary to task notes via `write_task_note`: what was built, the deliverable path(s), and a one-line confirmation that each criterion is met. Then report DONE.",
+      "suggestedRole": "developer",
+      "terminal": true
+    }
+  ],
+  "version": "1.0.2",
+  "releasedAt": "2026-08-21T00:00:00Z",
+  "paramSchema": {
+    "type": "object",
+    "properties": {
+      "workPath": {
+        "type": "string",
+        "title": "Working folder",
+        "description": "Per-task working folder in the artifacts drawer. Defaults to this task's own folder so runs never collide; override with a stable name when you deliberately want runs to share files.",
+        "default": "{{task.dir}}"
+      }
+    }
+  },
+  "minGezelVersion": "1.26233"
+}

package/data/craftbook-templates/in/insurance-inventory/versions/1.0.2/test.json

@@ -0,0 +1,124 @@
+{
+  "schemaVersion": 1,
+  "title": "Insurance Inventory smoke eval",
+  "objective": "Task-specific eval for the Insurance Inventory craftbook: a ten-item catalog becomes one agent-ready inventory document with every value traceable, the two unvalued items named as gaps, and a stated total.",
+  "tags": [
+    "data"
+  ],
+  "prompt": "our agent asked for a full inventry of the collection with values — can you put one together?",
+  "setup": {
+    "projectName": "Collection Catalog — Insurance Inventory Eval",
+    "about": "Self-contained eval project for insurance-inventory. The catalog store lives at collection.json in the workspace: ten items, two of them without a recorded value. The inventory document goes to insurance-inventory.md and its cover note to tasks/eval/report.md.",
+    "missionObjectives": "Use the Insurance Inventory craftbook/template, compile everything from the seeded collection.json, and finish with insurance-inventory.md plus tasks/eval/report.md — no network calls, real credentials, or live services.",
+    "files": [
+      {
+        "path": "collection.json",
+        "content": "[\n  { \"id\": \"v-001\", \"name\": \"John Coltrane - Blue Train\", \"category\": \"jazz-lp\", \"condition\": \"VG+\", \"photo\": \"photos/v-001.jpg\", \"estValueUsd\": 220, \"acquired\": \"2023-05-14, estate sale\" },\n  { \"id\": \"v-002\", \"name\": \"Nina Simone - I Put a Spell on You\", \"category\": \"vocal-jazz-lp\", \"condition\": \"VG\", \"photo\": \"photos/v-002.jpg\", \"estValueUsd\": 95, \"acquired\": \"2023-09-02, Vinyl Alley\" },\n  { \"id\": \"v-003\", \"name\": \"The Beatles - Abbey Road\", \"category\": \"rock-lp\", \"condition\": \"VG+\", \"photo\": \"photos/v-003.jpg\", \"estValueUsd\": 85, \"acquired\": \"2024-01-20, gift\" },\n  { \"id\": \"v-004\", \"name\": \"Fleetwood Mac - Rumours\", \"category\": \"rock-lp\", \"condition\": \"NM\", \"photo\": \"photos/v-004.jpg\", \"estValueUsd\": 60, \"acquired\": \"2024-03-08, Vinyl Alley\" },\n  { \"id\": \"v-005\", \"name\": \"Bill Evans - Waltz for Debby\", \"category\": \"jazz-lp\", \"condition\": \"VG\", \"photo\": \"photos/v-005.jpg\", \"estValueUsd\": 240, \"acquired\": \"2024-06-30, online auction\" },\n  { \"id\": \"v-006\", \"name\": \"Stevie Wonder - Innervisions\", \"category\": \"soul-lp\", \"condition\": \"VG+\", \"photo\": \"photos/v-006.jpg\", \"estValueUsd\": 55, \"acquired\": \"2024-10-12, flea market\" },\n  { \"id\": \"v-007\", \"name\": \"Dolly Parton - Jolene\", \"category\": \"country-lp\", \"condition\": \"VG\", \"photo\": \"photos/v-007.jpg\", \"estValueUsd\": 45, \"acquired\": \"2025-02-23, flea market\" },\n  { \"id\": \"v-008\", \"name\": \"Herbie Hancock - Head Hunters\", \"category\": \"jazz-lp\", \"condition\": \"VG+\", \"photo\": \"photos/v-008.jpg\", \"estValueUsd\": 70, \"acquired\": \"2025-07-19, Vinyl Alley\" },\n  { \"id\": \"v-009\", \"name\": \"Miles Davis - Kind of Blue\", \"category\": \"jazz-lp\", \"condition\": \"VG\", \"photo\": \"photos/v-009.jpg\", \"acquired\": \"2026-07-03, estate sale\" },\n  { \"id\": \"v-010\", \"name\": \"Joni Mitchell - Blue\", \"category\": \"folk-lp\", \"condition\": \"NM\", \"photo\": \"photos/v-010.jpg\", \"acquired\": \"2026-07-03, estate sale\" }\n]\n"
+      }
+    ],
+    "worker": {
+      "name": "Casper",
+      "role": "Curator"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "insurance-inventory.md holds a full ten-row table with values copied from the catalog, the unvalued items appear as named gaps, and the cover note states the recorded total of 870.",
+    "deliverables": [
+      {
+        "path": "insurance-inventory.md",
+        "kind": "markdown-doc",
+        "minBytes": 800,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "insurance-inventory.md",
+            "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+            "flags": "i"
+          },
+          {
+            "kind": "tableShape",
+            "file": "insurance-inventory.md",
+            "requiredColumns": [
+              "Item",
+              "Value",
+              "Photo"
+            ],
+            "minRows": 10
+          },
+          {
+            "kind": "contains",
+            "file": "insurance-inventory.md",
+            "pattern": "Kind of Blue",
+            "label": "the unvalued items still appear in the inventory"
+          },
+          {
+            "kind": "contains",
+            "file": "insurance-inventory.md",
+            "pattern": "missing|gap|no value|not (yet )?valued|unvalued",
+            "flags": "i",
+            "label": "value gaps named, not estimated away"
+          }
+        ]
+      },
+      {
+        "path": "tasks/eval/report.md",
+        "kind": "markdown-report",
+        "minBytes": 500,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "tasks/eval/report.md",
+            "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "tasks/eval/report.md",
+            "pattern": "\\b870\\b",
+            "label": "total of the recorded values stated correctly"
+          },
+          {
+            "kind": "contains",
+            "file": "tasks/eval/report.md",
+            "pattern": "missing|gap|no value|not (yet )?valued|unvalued",
+            "flags": "i",
+            "label": "cover note owns the gaps"
+          }
+        ]
+      }
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "insurance-inventory.md",
+      "kind": "markdown"
+    },
+    "axes": [
+      {
+        "name": "audit-readiness",
+        "description": "An insurance agent could use the document cold: full table, photo references, acquisition info where recorded."
+      },
+      {
+        "name": "conservation",
+        "description": "Every value traces to the catalog — no invented or rounded valuations, and the total re-adds."
+      },
+      {
+        "name": "gap-honesty",
+        "description": "Items without a value or photo are named with a next step, never quietly dropped or estimated."
+      },
+      {
+        "name": "presentation",
+        "description": "Clean structure and calm wording appropriate for a document that may back a claim."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "value traceability",
+    "gap disclosure",
+    "agent-ready presentation"
+  ]
+}