@bendyline/gilde 0.1.38 → 0.1.39

package/data/craftbook-templates/ac/accessibility-retrofit/versions/1.0.2/craftbook.json

@@ -0,0 +1,122 @@
+{
+  "id": "accessibility-retrofit",
+  "name": "Accessibility Retrofit (WCAG AA)",
+  "description": "Take an existing HTML page and retrofit it to WCAG 2.1 AA — fixing semantic structure, alt text, color contrast, keyboard operability, focus visibility, form labels, ARIA, and landmarks — without changing the visual design. Audits the page against the AA success criteria first to produce a prioritized findings list, then applies the fixes in place. Accessibility work fails when fixes are guessed rather than measured, so the audit phase enumerates the concrete violations and the target criteria before any edit, and the loop runs back to the FIX phase until every finding is resolved.\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. Audit (reviewer) — enumerate concrete WCAG-AA violations to fix → gated on artifact `{{workPath}}/audit.md` (markdown-notes)\n2. Fix (developer) — apply the a11y fixes in place, preserving design → gated on `index.html` (html-page)\n\nThe gates never advance with an unmet criterion, and loop back to the\nowning phase to fix named gaps.\n",
+  "entryStepId": "audit",
+  "triggers": [
+    "accessibility fix",
+    "make it accessible",
+    "wcag compliance",
+    "a11y retrofit",
+    "fix accessibility",
+    "screen reader friendly"
+  ],
+  "steps": [
+    {
+      "id": "audit",
+      "name": "Audit",
+      "description": "enumerate concrete WCAG-AA violations to fix",
+      "prompt": "Audit the existing page against WCAG 2.1 AA — do NOT fix yet. (1) Identify the target file and read it. (2) Check the high-yield AA criteria and record each concrete violation: missing/empty alt text on informative images, low text contrast (<4.5:1 normal, <3:1 large — estimate from the hex), missing form <label>s, non-semantic structure (divs where headings/landmarks/lists belong), keyboard traps or non-focusable interactive elements, missing focus indicator, missing lang attribute, missing skip link, improper heading order, and ARIA misuse. (3) For each finding note location, the failing criterion, and the fix. (4) Prioritize (blocker -> minor). (5) Write this as an acceptance-criteria checklist where each item is 'finding X is resolved' so evaluate can verify each. Write the findings + checklist to {{workPath}}/audit.md and via write_task_note. No edits yet.\n\nThe deliverable `{{workPath}}/audit.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",
+      "advanceWhen": {
+        "file": "{{workPath}}/audit.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/audit.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/audit.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "audit",
+        "maxAttempts": 3
+      },
+      "next": "fix"
+    },
+    {
+      "id": "fix",
+      "name": "Fix",
+      "description": "apply the a11y fixes in place, preserving design",
+      "prompt": "Apply the fixes to the page (write the corrected file as index.html) resolving every audited finding WITHOUT changing the visual design. (1) Add meaningful alt text (or alt=\"\" for decorative). (2) Adjust colors only as needed to hit AA contrast, staying on-brand. (3) Add <label>s / aria-label and associate them. (4) Use semantic HTML and landmarks (header/nav/main/footer), correct heading order, and lists. (5) Make all interactive elements keyboard-operable with a visible :focus indicator; add a skip link and the lang attribute; fix any ARIA. Change markup/CSS for semantics and a11y, not the look. On a loop-back, fix only the still-open findings. write_task_note the path and which findings are now resolved.",
+      "suggestedRole": "developer",
+      "advanceWhen": {
+        "file": "index.html",
+        "minBytes": 1,
+        "sniff": "html-complete"
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "index.html",
+            "bytes": 800
+          },
+          {
+            "kind": "htmlLint",
+            "file": "index.html"
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkHtmlComplete",
+            "scope": "standard",
+            "inputs": {
+              "file": "index.html"
+            }
+          }
+        ],
+        "onReject": "fix",
+        "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": "Verify the retrofit against the audit. Open/inspect index.html, tab through with the keyboard, and check contrast. Go through EACH finding in {{workPath}}/audit.md and write PASS/FAIL with a reason: is every image's alt correct, does text meet AA contrast, is every control labeled and keyboard-operable with a visible focus ring, is the structure semantic with landmarks and correct heading order, is there a lang attribute and skip link, and is the original visual design preserved? Any unresolved finding, or a visual regression, is a FAIL — loop back to fix.\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: \"fix\" })` 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": "index.html"
+        }
+      ],
+      "next": "fix"
+    },
+    {
+      "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/ac/accessibility-retrofit/versions/1.0.2/test.json

@@ -0,0 +1,100 @@
+{
+  "schemaVersion": 1,
+  "title": "Accessibility Retrofit (WCAG AA) smoke eval",
+  "objective": "Self-contained smoke eval for the Accessibility Retrofit (WCAG AA) craftbook using the html-page generic harness.",
+  "tags": [
+    "html-page"
+  ],
+  "prompt": "Can you biuld us a little page for this? The content notes are in source/page-content.md. Save it as index.html — one file, nothing fancy needed on our end.",
+  "setup": {
+    "projectName": "Accessibility Retrofit (WCAG AA) Eval",
+    "about": "Self-contained eval project for accessibility-retrofit. Seeded inputs are under workspace/source or workspace/fixtures; final deliverable is workspace/index.html.",
+    "missionObjectives": "Use the Accessibility Retrofit (WCAG AA) craftbook/template, read the seeded local fixtures, and write index.html without network calls, real credentials, or live services.",
+    "files": [
+      {
+        "path": "source/brief.md",
+        "content": "# Accessibility Retrofit (WCAG AA) 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: accessibility-retrofit - Accessibility Retrofit (WCAG AA).\n"
+      },
+      {
+        "path": "source/page-content.md",
+        "content": "# Boreal Desk Support Board\n\nBuild content around a support operations board. Required labels: Open issues, Status silence, Response time, Refund leakage, Next action. Required metric text: 18 hours to 6 hours and 14.2% to 8.9%.\n"
+      }
+    ],
+    "worker": {
+      "name": "Jules",
+      "role": "Developer"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "index.html is a source-grounded self-contained HTML page/tool.",
+    "deliverables": [
+      {
+        "path": "index.html",
+        "kind": "html-page",
+        "minBytes": 3000,
+        "checks": [
+          {
+            "kind": "cssMinBytes",
+            "bytes": 650,
+            "file": "index.html"
+          },
+          {
+            "kind": "jsParses",
+            "file": "index.html"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "Boreal|Support Board|returns",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "18\\s*hours|6\\s*hours|14\\.2%|8\\.9%",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "button|input|filter|tab|card|section|nav",
+            "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": [
+    "self-contained UI",
+    "source grounding",
+    "responsive structure"
+  ]
+}

package/data/craftbook-templates/ad/ad-variations/versions/1.0.2/craftbook.json

@@ -0,0 +1,117 @@
+{
+  "id": "ad-variations",
+  "name": "A/B Ad Copy Variations",
+  "description": "Generate a set of distinct, testable ad-copy variations for one offer across the angles worth A/B testing: it first locks the platform and its constraints (headline/description character limits), the offer, audience, and the distinct messaging angles to test (benefit, pain, social proof, urgency, curiosity), then drafts multiple complete variants per angle so each is a genuine test of a hypothesis rather than a trivial reword. Covers ad copy, A/B testing, Google/Facebook/LinkedIn ads, PPC copy, ad variations, and marketing experiments. Locking the angles and limits before writing is the point: a small model that commits to distinct hypotheses produces variants that actually test something, not ten near-duplicates that teach the advertiser nothing.\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. Ad brief (planner) — lock platform limits, offer, audience, and test angles → gated on artifact `{{workPath}}/brief.md` (markdown-notes)\n2. Draft the variations (copywriter) — write distinct, limit-respecting variants per angle → gated on `ad-variations.json` (json)\n\nThe gates never advance with an unmet criterion, and loop back to the\nowning phase to fix named gaps.\n",
+  "entryStepId": "brief",
+  "triggers": [
+    "write ad copy",
+    "ab test ad variations",
+    "facebook ad copy",
+    "google ads copy",
+    "ad variations"
+  ],
+  "steps": [
+    {
+      "id": "brief",
+      "name": "Ad brief",
+      "description": "lock platform limits, offer, audience, and test angles",
+      "prompt": "1) Pick the platform and its hard limits (e.g. Google RSA headlines <=30 chars and descriptions <=90 chars; Meta primary text, headline, description fields). 2) State the offer and the single primary audience. 3) Define the distinct ANGLES to test as named hypotheses (e.g. benefit-led, pain/problem-led, social-proof, urgency/scarcity, curiosity) — each a different reason the user clicks. 4) Decide how many variants per angle (e.g. 2) and the required fields per ad. 5) Note the CTA options. Write an acceptance-criteria checklist (e.g. 'every variant respects the platform char limits'; 'each angle is genuinely distinct, not a reword'; 'N variants per angle delivered'; 'every field populated'; 'one offer, consistent across variants'; 'CTA present'). Use write_task_note for the brief + angle list + checklist, and write the same to the produces path. No ad copy yet.\n\nThe deliverable `{{workPath}}/brief.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}}/brief.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/brief.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/brief.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "brief",
+        "maxAttempts": 3
+      },
+      "next": "draft"
+    },
+    {
+      "id": "draft",
+      "name": "Draft the variations",
+      "description": "write distinct, limit-respecting variants per angle",
+      "prompt": "Write the ad variations per the brief. 1) For each named angle, write the specified number of complete variants, each filling every required platform field (e.g. headline, description, primary text, CTA). 2) Make each angle genuinely distinct — a curiosity ad and a social-proof ad should read differently, not swap a synonym. 3) Strictly respect every character limit; count characters and stay under. 4) Keep the offer identical across all variants (only the angle/message changes). 5) Use punchy, concrete, scannable copy. 6) Output a single valid JSON array of objects: {\"angle\", \"variant\", \"headline\", \"description\", \"primaryText\" (if applicable), \"cta\", \"charCounts\"}. Save to the produces path and note total variant count.",
+      "suggestedRole": "copywriter",
+      "advanceWhen": {
+        "file": "ad-variations.json",
+        "minBytes": 1,
+        "sniff": "json-valid"
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "ad-variations.json",
+            "bytes": 2
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkJsonValid",
+            "scope": "standard",
+            "inputs": {
+              "file": "ad-variations.json"
+            }
+          }
+        ],
+        "onReject": "draft",
+        "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": "Parse ad-variations.json (must be valid JSON) and grade against the checklist. Check EACH criterion and write PASS/FAIL: (a) every field of every variant respects the platform's character limits (verify charCounts against the limits); (b) each angle is genuinely distinct from the others, not a reword; (c) the specified number of variants per angle is present; (d) every required field is populated; (e) the offer is consistent across all variants; (f) each has a CTA. A set of near-duplicate variants or any that exceeds a limit FAILS — loop back to draft and name the offenders.\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: \"draft\" })` 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": "ad-variations.json"
+        }
+      ],
+      "next": "draft"
+    },
+    {
+      "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/ad/ad-variations/versions/1.0.2/test.json

@@ -0,0 +1,111 @@
+{
+  "schemaVersion": 1,
+  "title": "Distinct ad-copy variations",
+  "objective": "Measure whether the ad-variations craftbook locks platform limits and produces genuinely distinct, complete ad variants instead of near-duplicate copy.",
+  "tags": [
+    "corpus"
+  ],
+  "prompt": "In the `Ad Variations Eval` project, write `ad-variations.json` for a Google search ad test for Harbor & Finch, a home-office accessories shop launching an ergonomic desk-lamp bundle. Use Google RSA-style limits: headlines <=30 characters and descriptions <=90 characters. Produce exactly 2 variants each for benefit-led, pain/problem-led, and social-proof angles. Include charCounts for every headline and description, and make each angle visibly distinct.",
+  "setup": {
+    "projectName": "Ad Variations Eval",
+    "about": "Self-contained eval project for ad-variations. The only deliverable is workspace/ad-variations.json.",
+    "missionObjectives": "Produce distinct, limit-aware ad variants for three named angles with complete fields and charCounts.",
+    "files": [],
+    "worker": {
+      "name": "Janne",
+      "role": "Copywriter"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "ad-variations.json is valid JSON with named angles, complete ad fields, and character counts.",
+    "deliverables": [
+      {
+        "path": "ad-variations.json",
+        "kind": "json",
+        "minBytes": 800,
+        "checks": [
+          {
+            "kind": "sniff",
+            "file": "ad-variations.json",
+            "sniff": "json-valid"
+          },
+          {
+            "kind": "contains",
+            "file": "ad-variations.json",
+            "pattern": "benefit-led|benefit",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "ad-variations.json",
+            "pattern": "pain|problem",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "ad-variations.json",
+            "pattern": "social-proof|social proof",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "ad-variations.json",
+            "pattern": "headline",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "ad-variations.json",
+            "pattern": "description",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "ad-variations.json",
+            "pattern": "char(acter)?[_ ]?counts?",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "ad-variations.json",
+            "pattern": "Harbor|Finch|desk-lamp|ergonomic",
+            "flags": "i"
+          }
+        ]
+      }
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "ad-variations.json",
+      "kind": "json"
+    },
+    "axes": [
+      {
+        "name": "grounding",
+        "description": "Seeded facts appear faithfully and the source is cited."
+      },
+      {
+        "name": "structure",
+        "description": "Sections progress logically for the artifact class."
+      },
+      {
+        "name": "completeness",
+        "description": "Required elements (risks, next actions, scope) are all present."
+      },
+      {
+        "name": "tone",
+        "description": "The register fits the audience and artifact class."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "angle distinctness",
+    "platform limits",
+    "structured JSON"
+  ]
+}

package/data/craftbook-templates/al/album-curate/versions/1.0.2/craftbook.json

@@ -0,0 +1,127 @@
+{
+  "id": "album-curate",
+  "name": "Curate an Album",
+  "description": "Turn a thousand camera-roll photos into the forty that tell the story: cull the near-duplicates, keep the keepers for honest reasons, caption what matters, and leave an album manifest the family will actually revisit. Locks the occasion and the selection criteria FIRST, then curates the manifest one moment at a time, then writes the album's short story. Use for curating an album, photo culling, 'make an album from the weekend', picking the keepers, and turning imports into albums.\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 album (archivist) — occasion, size, and the selection criteria locked → gated on artifact `{{workPath}}/scope.md` (markdown-notes)\n2. Curate the manifest (curator) — cluster the bursts, pick the best frames, write the manifest\n3. Write the album's story (copywriter) — the arc, the cuts worth mentioning, what to do next → 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": [
+    "curate an album",
+    "make an album from these photos",
+    "pick the keepers",
+    "cull the photos",
+    "album from the weekend"
+  ],
+  "steps": [
+    {
+      "id": "scope",
+      "name": "Scope the album",
+      "description": "occasion, size, and the selection criteria locked",
+      "prompt": "Define the album before picking a photo. Step 1: read photos/index.json from the workspace — the imported photos with their timestamps — and read any caption notes alongside them (photos/*.txt) describing what the camera saw. Step 2: confirm the occasion and size with the owner — what the album is about, roughly how many photos it should hold, and anything that must be in or out. Step 3: lock the selection criteria — near-duplicates and bursts resolve to the single best frame; every pick carries a caption and an honest reason it made the cut; the picks in order tell the occasion's story, start to end. Step 4: write an acceptance-criteria checklist ('album within the agreed size', 'every pick exists in the photo index — nothing invented', 'each duplicate cluster contributes at most one frame', 'every pick has a caption and a reason', 'the sequence reads as the story of the occasion'). Call write_task_note with the photo count + target size + 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": "archivist",
+      "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": "select"
+    },
+    {
+      "id": "select",
+      "name": "Curate the manifest",
+      "description": "cluster the bursts, pick the best frames, write the manifest",
+      "prompt": "Curate the album manifest. Step 1: walk the index in time order and group the near-duplicates — bursts and retakes shot seconds apart are one moment, not five photos. Step 2: from each moment worth keeping, choose the single best frame, using the caption notes where they exist to judge what the frame actually shows. Step 3: order the picks so the album tells the occasion start to end. Step 4: write the album manifest into albums/ named for the occasion — a JSON array, one record per pick with file, caption, and reason, the file spelled exactly as the index spells it. Call write_task_note with the pick count and the moments you cut.",
+      "suggestedRole": "curator",
+      "next": "report"
+    },
+    {
+      "id": "report",
+      "name": "Write the album's story",
+      "description": "the arc, the cuts worth mentioning, what to do next",
+      "prompt": "Write the album's short story for whoever opens it. Step 1: the arc — what the occasion was and how the picks carry it, in a paragraph. Step 2: the cuts worth mentioning — the burst that became one frame, the beloved-but-blurry shot that did not make it — so the family knows the culling was care, not loss. Step 3: anything the owner should re-shoot, print, or ask about. Keep it warm and short. Write {{workPath}}/report.md and call write_task_note with the album path and the pick count.\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 the album manifest in albums/, {{workPath}}/report.md, and photos/index.json and verify every criterion from {{workPath}}/scope.md: the album is within the agreed size, every pick's file appears in the photo index exactly (spot-check several — no invented or misspelled files), no duplicate cluster contributes more than one frame (check the picks' timestamps), every record carries a nonempty caption and reason, and the order tells the story. Write PASS/FAIL per criterion; on any failure, name the gap and loop back to select.\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: \"select\" })` 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": "select"
+    },
+    {
+      "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"
+}