@bendyline/gilde 0.1.38 → 0.1.39

package/data/craftbook-templates/co/comparison-matrix/versions/1.0.2/test.json

@@ -0,0 +1,112 @@
+{
+  "schemaVersion": 1,
+  "title": "Weighted vendor comparison matrix",
+  "objective": "Measure whether the comparison-matrix craftbook locks criteria and builds an interactive scored matrix from seeded option facts.",
+  "tags": [
+    "html-page"
+  ],
+  "prompt": "In the `Comparison Matrix Eval` project, read `source/options.md` and build `index.html`: an interactive weighted comparison for choosing a returns-management vendor. Compare Returnly, LoopDesk, and ParcelPilot across implementation effort, automation, reporting, and total cost. Show weights, scores, totals, a highlighted recommendation, and sortable table behavior. Vanilla HTML/CSS/JS only.",
+  "setup": {
+    "projectName": "Comparison Matrix Eval",
+    "about": "Self-contained eval project for the comparison-matrix craftbook. Option facts are in workspace/source/options.md.",
+    "missionObjectives": "Build a scored vendor comparison page with locked criteria weights, complete option cells, computed totals, recommendation, and interactive sorting.",
+    "files": [
+      {
+        "path": "source/options.md",
+        "content": "# Returns Vendor Options\n\nGoal: choose a returns-management vendor for a 90-day pilot.\nCriteria weights: implementation effort 25%, automation 30%, reporting 25%, total cost 20%.\n\n## Returnly\n- Implementation effort: medium, 4 weeks.\n- Automation: strong label rules and refund triggers.\n- Reporting: basic CSV export.\n- Total cost: $4,800 pilot.\n\n## LoopDesk\n- Implementation effort: low, 2 weeks.\n- Automation: moderate intake routing.\n- Reporting: strong dashboard and exception export.\n- Total cost: $5,600 pilot.\n\n## ParcelPilot\n- Implementation effort: high, 6 weeks.\n- Automation: strongest warehouse barcode workflow.\n- Reporting: moderate canned reports.\n- Total cost: $4,200 pilot.\n"
+      }
+    ],
+    "worker": {
+      "name": "Daan",
+      "role": "Developer"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "workspace/index.html is a complete sortable comparison page with weights, totals, all options, and a recommendation.",
+    "deliverables": [
+      {
+        "path": "index.html",
+        "kind": "html-page",
+        "minBytes": 3600,
+        "checks": [
+          {
+            "kind": "cssMinBytes",
+            "bytes": 650,
+            "file": "index.html"
+          },
+          {
+            "kind": "jsParses",
+            "file": "index.html"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "Returnly|LoopDesk|ParcelPilot"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "25%|30%|20%"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "implementation effort|automation|reporting|total cost",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "score|weighted|total",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "recommend|best choice",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "sort|addEventListener|onclick",
+            "flags": "i"
+          }
+        ]
+      }
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "index.html",
+      "kind": "html"
+    },
+    "axes": [
+      {
+        "name": "structure",
+        "description": "Layout and hierarchy fit the content; sections are scannable and coherent."
+      },
+      {
+        "name": "grounding",
+        "description": "Seeded source content is used faithfully — no invented facts or dropped requirements."
+      },
+      {
+        "name": "interactivity",
+        "description": "Controls and states the page claims to offer actually work."
+      },
+      {
+        "name": "polish",
+        "description": "Styling is intentional and responsive rather than unstyled or broken."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "criteria weighting",
+    "complete matrix",
+    "computed recommendation"
+  ]
+}

package/data/craftbook-templates/co/compat-check/versions/1.0.2/craftbook.json

@@ -0,0 +1,158 @@
+{
+  "id": "compat-check",
+  "name": "Compatibility Check",
+  "description": "Answer 'does it still work on the new version?' with a matrix, not a shrug: the datapack's files checked against each target version's format rules, every break named with the exact file and the exact fix, and the upgrade path written down in order. Locks the target version list and the compat criteria FIRST (format numbers, renamed folders, changed identifiers), then checks version by version, then writes the matrix and the fixes. Use for version compatibility checks, 'does the mod work on 1.21', update breakage sweeps, and upgrade planning.\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 check (modmaker) — target versions, format rules, compat criteria locked → gated on artifact `{{workPath}}/scope.md` (markdown-notes)\n2. Check each version (developer) — the pack walked against each version's rules, findings recorded → gated on artifact `{{workPath}}/checks.md` (markdown-notes)\n3. Write the matrix (reviewer) — the compat matrix, the fixes, the ordered upgrade path → 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": [
+    "check compatibility",
+    "does it work on the new version",
+    "will the mod break",
+    "version check the datapack",
+    "plan the upgrade"
+  ],
+  "steps": [
+    {
+      "id": "scope",
+      "name": "Scope the check",
+      "description": "target versions, format rules, compat criteria locked",
+      "prompt": "Define the check before judging a single file. Step 1: list the target game versions to check against, and read the version notes in the workspace (the version-to-pack_format table and any renamed-folder or changed-identifier notes) — every claim in the matrix must trace to these notes or to the pack's own files, never to memory. Step 2: inventory the datapack — pack.mcmeta's current pack_format, the folder names in use under data/, and the identifiers the functions and JSON files rely on. Step 3: lock the compat criteria per version — the pack_format the version expects, the folder names it expects, and the identifiers it renamed or removed; anything the notes do not cover gets checked as 'unverified', not guessed. Step 4: write an acceptance-criteria checklist ('every target version gets a verdict', 'every break names the exact file and what breaks', 'every break has a concrete fix', 'verdicts trace to the version notes or the pack files', 'the upgrade path is ordered'). Call write_task_note with the version list + criteria + 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": "modmaker",
+      "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": "check"
+    },
+    {
+      "id": "check",
+      "name": "Check each version",
+      "description": "the pack walked against each version's rules, findings recorded",
+      "prompt": "Walk the datapack against each target version's locked criteria. Step 1: for each version, check pack.mcmeta's pack_format against the version's expected number; a mismatch is a break with a one-line fix. Step 2: check every folder under data/ against the version's expected folder names — renamed folders are the silent killer, the pack loads and the content just never fires. Step 3: sweep the function and JSON files for identifiers the version renamed or removed, listing each occurrence by file. Step 4: record the findings per version in {{workPath}}/checks.md — works, breaks with the named file and reason, or unverified where the notes ran out — and be honest about the unverified ones; a confident wrong verdict costs an evening of debugging. Call write_task_note with the per-version verdict counts.\n\nThe deliverable `{{workPath}}/checks.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": "developer",
+      "advanceWhen": {
+        "file": "{{workPath}}/checks.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/checks.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/checks.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "check",
+        "maxAttempts": 3
+      },
+      "next": "report"
+    },
+    {
+      "id": "report",
+      "name": "Write the matrix",
+      "description": "the compat matrix, the fixes, the ordered upgrade path",
+      "prompt": "Write the report that answers the question at a glance. Step 1: open with the verdict in one sentence — which versions the pack works on today and which break it. Step 2: build the compatibility matrix as a table — a row per target version with columns for Version and Status at minimum, the status stated plainly (works, breaks, unverified) with the one-line reason. Step 3: list the fixes — every break from {{workPath}}/checks.md with the exact file and the exact change, grouped by version. Step 4: write the upgrade path — the fixes in the order to apply them, and what to test in-game after each, so the upgrade is an afternoon of steps rather than a weekend of guessing. Write {{workPath}}/report.md and call write_task_note with the report path and the verdict sentence.\n\nBefore working, open the artifact input `{{workPath}}/checks.md` with `read_artifact`; do not look for it in the workspace.\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": "reviewer",
+      "consumes": [
+        {
+          "file": "{{workPath}}/checks.md",
+          "artifact": true
+        }
+      ],
+      "advanceWhen": {
+        "file": "{{workPath}}/report.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/report.md",
+            "bytes": 1000,
+            "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 {{workPath}}/report.md, {{workPath}}/checks.md, and the version notes and verify every criterion from {{workPath}}/scope.md: every target version has a row in the matrix with a plain verdict, every break names the exact file and carries a concrete fix, every verdict traces to the version notes or the pack's actual files (spot-check two against the notes), unverified items are marked as such rather than guessed, and the upgrade path is ordered with a test step per fix. Write PASS/FAIL per criterion; on any failure, name the gap and loop back to report.\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: \"report\" })` 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": "report"
+    },
+    {
+      "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/co/compat-check/versions/1.0.2/test.json

@@ -0,0 +1,148 @@
+{
+  "schemaVersion": 1,
+  "title": "Compatibility Check smoke eval",
+  "objective": "Task-specific eval for the Compatibility Check craftbook: a 1.20-era boat-diamond datapack (pack_format 15, plural folder names) checked against 1.21.5 and 1.20.6 using the house version notes — a Version/Status matrix, every break named with its file and fix, and an ordered upgrade path.",
+  "tags": [
+    "corpus"
+  ],
+  "prompt": "does our boat-diamond pack still work on 1.21.5? and what about 1.20.6? check compatability and write it up",
+  "setup": {
+    "projectName": "Mod Workshop — Compat Check Eval",
+    "about": "Self-contained eval project for compat-check. The datapack under test sits in the workspace: pack.mcmeta (currently pack_format 15) and the boatpack content under data/ using the pre-1.21 plural folder names. The house version notes at tasks/eval/format-versions.md carry the pack_format table and the 1.21 folder-rename rule. Findings go to tasks/eval/checks.md and the matrix report to tasks/eval/report.md.",
+    "missionObjectives": "Use the Compatibility Check craftbook/template, check the pack against 1.21.5 and 1.20.6 using only the version notes and the pack's own files, and write tasks/eval/report.md with a Version/Status matrix, per-break fixes, and an ordered upgrade path. Claims trace to the notes or the files — anything the notes do not cover is marked unverified, never guessed.",
+    "files": [
+      {
+        "path": "tasks/eval/format-versions.md",
+        "content": "# Datapack format numbers (house notes)\n\n| Game version | pack_format |\n| --- | --- |\n| 1.20 | 15 |\n| 1.20.6 | 41 |\n| 1.21 | 48 |\n| 1.21.5 | 71 |\n\nFolder names: since 1.21 the data folder names are singular - function,\nrecipe, advancement, loot_table, and tags/function. Before 1.21 they were\nplural - functions, recipes, advancements, loot_tables, tags/functions.\nA pack using the old plural names loads on 1.21+ but its content silently\nnever fires.\n"
+      },
+      {
+        "path": "pack.mcmeta",
+        "content": "{\n  \"pack\": {\n    \"pack_format\": 15,\n    \"description\": \"Boat Diamond - craft a boat, get a diamond\"\n  }\n}\n"
+      },
+      {
+        "path": "data/boatpack/functions/give_diamond.mcfunction",
+        "content": "give @s minecraft:diamond 1\nadvancement revoke @s only boatpack:craft_boat\ntellraw @s {\"text\":\"A boat! Have a diamond.\",\"color\":\"aqua\"}\n"
+      },
+      {
+        "path": "data/boatpack/advancements/craft_boat.json",
+        "content": "{\n  \"criteria\": {\n    \"crafted_boat\": {\n      \"trigger\": \"minecraft:inventory_changed\",\n      \"conditions\": {\n        \"items\": [\n          {\n            \"items\": [\n              \"minecraft:oak_boat\"\n            ]\n          }\n        ]\n      }\n    }\n  },\n  \"rewards\": {\n    \"function\": \"boatpack:give_diamond\"\n  }\n}\n"
+      }
+    ],
+    "worker": {
+      "name": "Jesse",
+      "role": "Developer"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "tasks/eval/report.md answers at a glance: a Version/Status matrix covering both asked versions, the pack_format mismatches named with the numbers from the notes, the 1.21 folder renames caught as breaks with exact files, and an ordered upgrade path.",
+    "deliverables": [
+      {
+        "path": "tasks/eval/report.md",
+        "kind": "markdown-report",
+        "minBytes": 800,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "tasks/eval/report.md",
+            "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+            "flags": "i"
+          },
+          {
+            "kind": "tableShape",
+            "file": "tasks/eval/report.md",
+            "requiredColumns": [
+              "Version",
+              "Status"
+            ],
+            "minRows": 2
+          },
+          {
+            "kind": "contains",
+            "file": "tasks/eval/report.md",
+            "pattern": "1\\.21\\.5",
+            "label": "the 1.21.5 verdict is present"
+          },
+          {
+            "kind": "contains",
+            "file": "tasks/eval/report.md",
+            "pattern": "1\\.20\\.6",
+            "label": "the 1.20.6 verdict is present"
+          },
+          {
+            "kind": "contains",
+            "file": "tasks/eval/report.md",
+            "pattern": "pack_format|format number",
+            "flags": "i",
+            "label": "the format-number break is named"
+          },
+          {
+            "kind": "contains",
+            "file": "tasks/eval/report.md",
+            "pattern": "folder|functions|advancements",
+            "flags": "i",
+            "label": "the renamed-folder break is named"
+          },
+          {
+            "kind": "valuesSubsetOf",
+            "file": "tasks/eval/report.md",
+            "sourceFiles": [
+              "tasks/eval/format-versions.md"
+            ],
+            "pattern": "\\b(1\\.2\\d(?:\\.\\d+)?)\\b",
+            "minMatches": 2,
+            "label": "every game version cited comes from the version notes"
+          },
+          {
+            "kind": "judge",
+            "file": "tasks/eval/report.md",
+            "rubric": "Each verdict traces to the version notes or the pack's actual files (pack_format 15 versus the required 71 and 41; the plural functions/advancements folders breaking on 1.21+), fixes name the exact file and change, the upgrade path is ordered with a test step per fix, and nothing outside the notes is asserted with confidence.",
+            "sourceFiles": [
+              "tasks/eval/format-versions.md",
+              "pack.mcmeta",
+              "tasks/eval/checks.md"
+            ],
+            "label": "grounded verdicts and concrete fixes"
+          }
+        ]
+      },
+      {
+        "path": "tasks/eval/checks.md",
+        "kind": "markdown-notes",
+        "minBytes": 300
+      }
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "tasks/eval/report.md",
+      "kind": "markdown"
+    },
+    "axes": [
+      {
+        "name": "thoroughness",
+        "description": "Both asked versions get real verdicts; format number and folder names are both checked."
+      },
+      {
+        "name": "precision",
+        "description": "Breaks name the exact file and the exact expected value or name."
+      },
+      {
+        "name": "upgrade-guidance",
+        "description": "The fixes are ordered into a path with a test step per change."
+      },
+      {
+        "name": "honesty",
+        "description": "Anything beyond the notes is marked unverified, never asserted from memory."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "grounded verdicts",
+    "exact-file fixes",
+    "ordered upgrade path"
+  ]
+}

package/data/craftbook-templates/co/competitive-analysis/versions/1.0.2/craftbook.json

@@ -0,0 +1,152 @@
+{
+  "id": "competitive-analysis",
+  "name": "Competitive Analysis",
+  "description": "Compare a set of competitors across consistent dimensions and turn the comparison into an actionable analysis: a planner locks the competitor set and the evaluation dimensions, a researcher fills a feature/positioning matrix with evidence-backed cells, and a copywriter writes the analysis with a clear recommendation and competitive gaps. The ordering matters because dimensions chosen after the research are cherry-picked, and a write-up without a complete matrix has holes a reader will spot; locking the same dimensions for every competitor up front forces an apples-to-apples comparison. Use this for competitor teardowns, market positioning, win/loss analysis, 'how do we compare to X and Y' questions, and feature/pricing comparisons across rivals.\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. Dimension scope (planner) — lock the competitor set + comparison dimensions → gated on artifact `{{workPath}}/dimension-scope.md` (markdown-notes)\n2. Fill the matrix (researcher) — evidence-backed cell for every competitor x dimension → gated on artifact `{{workPath}}/matrix.md` (markdown-notes)\n3. Write the analysis (copywriter) — narrative analysis + recommendation from the matrix → gated on artifact `{{workPath}}/analysis.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": "dimension-scope",
+  "triggers": [
+    "competitive analysis",
+    "compare competitors",
+    "how do we stack up against",
+    "competitor teardown",
+    "market positioning analysis"
+  ],
+  "steps": [
+    {
+      "id": "dimension-scope",
+      "name": "Dimension scope",
+      "description": "lock the competitor set + comparison dimensions",
+      "prompt": "1. Identify the subject (us/the product) and pick 3-6 named competitors that are genuinely comparable; justify each inclusion in one line. 2. Choose 5-8 comparison dimensions that decide the market (e.g. core features, pricing model, target segment, differentiator, weaknesses, traction signals) and define what each dimension means so cells are comparable. 3. Decide the matrix schema (rows = competitors incl. us, columns = dimensions) and where the final analysis lands. 4. Write an acceptance-criteria checklist (e.g. 'every competitor x every dimension cell is filled with evidence, not blank or guessed', 'our own offering is included as a row', 'each cell cites or names a source', 'the analysis ends with a recommendation and the top competitive gaps'). 5. Record decisions. Call write_task_note with the competitor set, dimensions, schema, and checklist; write the same to {{workPath}}/dimension-scope.md.\n\nThe deliverable `{{workPath}}/dimension-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": "planner",
+      "advanceWhen": {
+        "file": "{{workPath}}/dimension-scope.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/dimension-scope.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/dimension-scope.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "dimension-scope",
+        "maxAttempts": 3
+      },
+      "next": "research"
+    },
+    {
+      "id": "research",
+      "name": "Fill the matrix",
+      "description": "evidence-backed cell for every competitor x dimension",
+      "prompt": "1. For each competitor, research every locked dimension and fill the corresponding cell in {{workPath}}/matrix.md with a specific, evidence-backed value (a feature fact, a real price, a named segment) — never 'maybe' or a guess. 2. Cite or name the source for each cell (product page, pricing page, review, filing) in a [n] reference list. 3. Use consistent wording per dimension across competitors so cells are directly comparable. 4. Where a value is genuinely unknown, mark the cell UNKNOWN and say why, rather than fabricating. 5. Note any standout strength or weakness you observe per competitor for the writer to use. Good looks like a fully-populated, cited matrix. Do not write the narrative analysis yet.\n\nThe deliverable `{{workPath}}/matrix.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}}/matrix.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/matrix.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/matrix.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "research",
+        "maxAttempts": 3
+      },
+      "next": "write"
+    },
+    {
+      "id": "write",
+      "name": "Write the analysis",
+      "description": "narrative analysis + recommendation from the matrix",
+      "prompt": "1. Write {{workPath}}/analysis.md: start with the rendered comparison matrix (a markdown table of competitors x dimensions), then a per-dimension or per-competitor narrative interpreting the matrix. 2. Explicitly call out where each competitor leads, where they are weak, and where the market has a gap. 3. End with a clear, defensible recommendation (where to compete, what to build/avoid) tied to specific cells in the matrix. 4. Keep every claim traceable to a matrix cell and its source; carry the [n] references through. 5. On a loop-back, fix only the named gaps. 6. Call write_task_note with the analysis path and which criteria now pass.\n\nThe deliverable `{{workPath}}/analysis.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}}/analysis.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/analysis.md",
+            "bytes": 1500,
+            "artifact": true
+          },
+          {
+            "kind": "contains",
+            "file": "{{workPath}}/analysis.md",
+            "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+            "flags": "i",
+            "label": "at least one markdown heading",
+            "artifact": true
+          }
+        ],
+        "onReject": "write",
+        "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}}/analysis.md and check EACH criterion PASS/FAIL: (1) the matrix table covers every competitor x every dimension from the scope with no unexplained blanks; (2) our own offering is included; (3) cells are evidence-backed and consistently worded, with sources named (spot-check 3 against {{workPath}}/matrix.md); (4) the narrative actually interprets the matrix rather than restating it; (5) there is a clear recommendation tied to specific evidence; (6) competitive gaps are explicitly identified. A pretty write-up over a half-empty or guessed matrix FAILS.\n\nThe deliverable lives in the project's artifacts drawer — open `{{workPath}}/analysis.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}}/analysis.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"
+}