@bendyline/gilde 0.1.38 → 0.1.39

package/data/craftbook-templates/ba/backup-routine/versions/1.0.2/test.json

@@ -0,0 +1,111 @@
+{
+  "schemaVersion": 1,
+  "title": "Cron-friendly backup script from retention brief",
+  "objective": "Measure whether the backup-routine craftbook scopes retention and integrity before producing a cautious backup script.",
+  "tags": [
+    "data"
+  ],
+  "prompt": "In the `Backup Routine Eval` project, read `source/backup-brief.md` and write `backup.sh`: a cron-friendly shell script that snapshots the source directory, writes a timestamped compressed archive, computes a checksum, verifies integrity, prunes old backups safely, logs each run, and exits non-zero on missing or empty sources.",
+  "setup": {
+    "projectName": "Backup Routine Eval",
+    "about": "Self-contained eval project for backup-routine. The operational brief is in workspace/source/backup-brief.md and the deliverable is workspace/backup.sh.",
+    "missionObjectives": "Write a cautious backup script with retention, integrity checking, newest-backup protection, logging, and explicit failure handling.",
+    "files": [
+      {
+        "path": "source/backup-brief.md",
+        "content": "# Backup routine brief\n\nSource directory: ./var/app-data. Exclude cache/ and tmp/. Destination: ./backups. Archive name: app-data-YYYYMMDD-HHMMSS.tar.gz. Retention: keep the last 7 verified backups and never delete the newest verified backup. Integrity: write a SHA-256 checksum file and verify the archive can be listed before marking it verified. Log file: backup.log with timestamp, archive path, size, checksum, verified flag, and pruned files. Failure handling: missing or empty source must abort non-zero and must not write a hollow archive.\n"
+      }
+    ],
+    "worker": {
+      "name": "Daan",
+      "role": "Developer"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "workspace/backup.sh is a runnable-looking shell routine with timestamped archive, checksum verification, retention pruning, logging, and guarded failure behavior.",
+    "deliverables": [
+      {
+        "path": "backup.sh",
+        "kind": "generic-file",
+        "minBytes": 1600,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "backup.sh",
+            "pattern": "set -euo pipefail|set -eu",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "backup.sh",
+            "pattern": "tar|gzip|\\.tar\\.gz",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "backup.sh",
+            "pattern": "sha256sum|shasum",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "backup.sh",
+            "pattern": "retention|keep|prune|delete",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "backup.sh",
+            "pattern": "newest|latest|verified",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "backup.sh",
+            "pattern": "backup.log|log",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "backup.sh",
+            "pattern": "exit 1|return 1|non-zero|missing|empty",
+            "flags": "i"
+          }
+        ]
+      }
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "backup.sh",
+      "kind": "text"
+    },
+    "axes": [
+      {
+        "name": "accuracy",
+        "description": "Computed figures are correct against the seeded records."
+      },
+      {
+        "name": "grounding",
+        "description": "Every claim traces to the seeded data — nothing invented."
+      },
+      {
+        "name": "structure",
+        "description": "Tables and sections make the analysis checkable at a glance."
+      },
+      {
+        "name": "actionability",
+        "description": "Recommendations follow from the findings and are concretely doable."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "integrity check",
+    "safe retention",
+    "non-zero failure handling"
+  ]
+}

package/data/craftbook-templates/bl/blog-post/versions/1.0.2/craftbook.json

@@ -0,0 +1,155 @@
+{
+  "id": "blog-post",
+  "name": "SEO Blog Post",
+  "description": "Write a complete, SEO-aware blog post that ranks and reads well: it locks the primary keyword, search intent, target reader, and a heading-by-heading outline FIRST, then drafts the full article in a single human voice with the keyword woven naturally into the title, intro, and H2s, and finally runs an editorial pass for clarity, flow, factual accuracy, and on-page SEO. Covers blog article, SEO content, how-to guide, listicle, thought-leadership post, content marketing, and organic-traffic writing. The scope-first ordering is the point: a small model that picks the keyword and outline before drafting avoids the classic failure of a rambling post that buries its topic and never satisfies the searcher's intent.\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. Keyword & outline scope (planner) — lock the primary keyword, intent, reader, and H2 outline → gated on artifact `{{workPath}}/keyword-scope.md` (markdown-notes)\n2. Draft the article (copywriter) — write the full post to the locked outline and keyword plan → gated on `blog-post.md` (markdown-doc)\n3. Editorial pass (reviewer) — tighten clarity, flow, accuracy, and on-page SEO → gated on `blog-post.md` (markdown-doc)\n\nThe gates never advance with an unmet criterion, and loop back to the\nowning phase to fix named gaps.\n",
+  "entryStepId": "keyword-scope",
+  "triggers": [
+    "write a blog post",
+    "seo article",
+    "write a how-to guide",
+    "create blog content",
+    "ranking article"
+  ],
+  "steps": [
+    {
+      "id": "keyword-scope",
+      "name": "Keyword & outline scope",
+      "description": "lock the primary keyword, intent, reader, and H2 outline",
+      "prompt": "1) Identify the PRIMARY keyword/phrase the post targets and the matching search intent (informational, comparison, how-to, or commercial). 2) Define the target reader in one sentence (who they are, what they already know, what they want to leave with). 3) List 3-5 secondary keywords/related terms to weave in naturally. 4) Write the working title (must contain the primary keyword, ideally near the front) plus a 1-sentence meta description under 155 characters. 5) Produce a heading-by-heading outline: H1 (title), a hook intro, 4-7 H2 sections each with a one-line summary of what it covers, and a closing/CTA section. 6) Set the target word count (e.g. 1200-1600) and reading level. Write an acceptance-criteria checklist (e.g. 'primary keyword appears in title, first 100 words, and at least 2 H2s'; 'every outlined H2 is covered'; 'single consistent voice, no AI filler phrases like \"in today's fast-paced world\"'; 'ends with a clear CTA'; 'within target word count +/-15%'). Use write_task_note to record the keyword strategy, outline, and checklist, and write the same content to the produces path. No prose drafting yet.\n\nThe deliverable `{{workPath}}/keyword-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}}/keyword-scope.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/keyword-scope.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/keyword-scope.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "keyword-scope",
+        "maxAttempts": 3
+      },
+      "next": "draft"
+    },
+    {
+      "id": "draft",
+      "name": "Draft the article",
+      "description": "write the full post to the locked outline and keyword plan",
+      "prompt": "Write the complete article to the locked outline. 1) Open with a hook intro (2-3 sentences) that names the reader's problem and previews the payoff; place the primary keyword in the first 100 words. 2) Write every outlined H2 section in full prose with concrete examples, short paragraphs (2-4 sentences), and lists where they aid scanning. 3) Weave secondary keywords in naturally — never keyword-stuff. 4) Use one consistent, specific voice; cut filler and hedging. 5) Close with the CTA section from the outline. 6) Output GitHub-flavored markdown with the H1 title at top, proper H2/H3 headers, and the meta description as an HTML comment or front-matter line at the top. Write the full post to the produces path and use write_task_note to record word count and which checklist items you believe pass.",
+      "suggestedRole": "copywriter",
+      "advanceWhen": {
+        "file": "blog-post.md",
+        "minBytes": 1,
+        "sniff": "nonempty"
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "blog-post.md",
+            "bytes": 500
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkContains",
+            "scope": "standard",
+            "inputs": {
+              "file": "blog-post.md",
+              "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+              "flags": "i"
+            }
+          }
+        ],
+        "onReject": "draft",
+        "maxAttempts": 3
+      },
+      "next": "edit"
+    },
+    {
+      "id": "edit",
+      "name": "Editorial pass",
+      "description": "tighten clarity, flow, accuracy, and on-page SEO",
+      "prompt": "Edit the draft in place against the checklist. 1) Read top to bottom for flow: does each section follow logically, are transitions smooth, is the voice consistent? 2) Cut every filler phrase, redundancy, and hedge; tighten long sentences. 3) Verify on-page SEO: primary keyword in title, first 100 words, and >=2 H2s; meta description present and under 155 chars; headings are descriptive. 4) Fact-check any concrete claim and flag/remove anything you cannot stand behind. 5) Confirm word count is within target and the CTA is clear. Rewrite weak passages directly rather than just commenting. Save the polished version back to the produces path and note any criteria still unmet.",
+      "suggestedRole": "reviewer",
+      "advanceWhen": {
+        "file": "blog-post.md",
+        "minBytes": 1,
+        "sniff": "nonempty"
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "blog-post.md",
+            "bytes": 500
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkContains",
+            "scope": "standard",
+            "inputs": {
+              "file": "blog-post.md",
+              "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+              "flags": "i"
+            }
+          }
+        ],
+        "onReject": "edit",
+        "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": "Read the final blog-post.md end to end and grade against the locked checklist. Check EACH criterion and write PASS/FAIL with a one-line reason: (a) primary keyword in title, first 100 words, and >=2 H2s; (b) every outlined H2 is present and substantive; (c) single consistent voice with no AI-filler phrases; (d) meta description present and under 155 characters; (e) clear CTA at the end; (f) word count within target +/-15%; (g) no unsupported factual claims. A post that is fluent but ignores the keyword plan or drops outlined sections FAILS.\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": "blog-post.md"
+        }
+      ],
+      "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/bl/blog-post/versions/1.0.2/test.json

@@ -0,0 +1,121 @@
+{
+  "schemaVersion": 1,
+  "title": "SEO-aware how-to blog post",
+  "objective": "Measure whether the blog-post craftbook scopes keyword intent and outline before producing a complete, SEO-aware article grounded in seeded product notes.",
+  "tags": [
+    "corpus"
+  ],
+  "prompt": "In the `Blog Post Eval` project, read `source/brief.md` and write `blog-post.md`: a how-to article targeting the primary keyword \"sustainable packaging checklist\" for operations managers at small ecommerce brands. It needs a meta description, H1 title, at least five H2 sections, practical checklist content, the seeded Harbor & Finch facts, and a clear CTA. Do not invent statistics beyond the source notes.",
+  "setup": {
+    "projectName": "Blog Post Eval",
+    "about": "Self-contained eval project for the blog-post craftbook. Source notes are in workspace/source/brief.md and the deliverable is workspace/blog-post.md.",
+    "missionObjectives": "Produce a scoped, SEO-aware markdown article with meta description, H1/H2 structure, source-grounded examples, checklist advice, and CTA.",
+    "files": [
+      {
+        "path": "source/brief.md",
+        "content": "# Sustainable Packaging Brief\n\nAudience: operations managers at ecommerce brands shipping 1,000-15,000 parcels per month.\nPrimary keyword: sustainable packaging checklist.\nReader intent: practical how-to before choosing packaging vendors.\nSeed facts:\n- Harbor & Finch cut filler material by 32% after right-sizing mailers.\n- Their damage rate stayed under 1.8% during a 60-day pilot.\n- Warehouse staff need a packing-station card, not a long policy PDF.\n- Finance wants packaging cost tracked per shipped order.\n- The CTA should invite teams to run a two-week packaging audit.\n"
+      }
+    ],
+    "worker": {
+      "name": "Nora",
+      "role": "Writer"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "workspace/blog-post.md is a complete markdown article with SEO metadata, H1/H2 structure, seeded facts, checklist guidance, and CTA.",
+    "deliverables": [
+      {
+        "path": "blog-post.md",
+        "kind": "markdown-doc",
+        "minBytes": 2200,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "blog-post.md",
+            "pattern": "meta description",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "blog-post.md",
+            "pattern": "^#\\s+",
+            "flags": "im"
+          },
+          {
+            "kind": "contains",
+            "file": "blog-post.md",
+            "pattern": "sustainable packaging checklist",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "blog-post.md",
+            "pattern": "##\\s+",
+            "flags": "m"
+          },
+          {
+            "kind": "contains",
+            "file": "blog-post.md",
+            "pattern": "32%|1\\.8%|60-day",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "blog-post.md",
+            "pattern": "packing-station card|cost tracked per shipped order",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "blog-post.md",
+            "pattern": "two-week packaging audit|CTA|call",
+            "flags": "i"
+          }
+        ]
+      }
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "blog-post.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": [
+    "keyword scope",
+    "outline-first writing",
+    "source-grounded article"
+  ],
+  "extensions": {
+    "legacySimulators": [
+      {
+        "id": "blog-post-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/bo/board-deck/versions/1.0.2/craftbook.json

@@ -0,0 +1,153 @@
+{
+  "id": "board-deck",
+  "name": "Board Meeting Deck",
+  "description": "Build a board-meeting deck that briefs directors efficiently — an executive summary up top, KPI dashboard, financials, progress against goals, key risks/asks, and an appendix — structured so a director gets the state of the business in minutes. Locks the standard board narrative and the metrics that matter FIRST (boards want a consistent dashboard, not a redesign every quarter), designs a dense data-forward slide system, then builds a navigable HTML deck where every slide leads with its number and the asks are unmistakable. Use for quarterly board meetings, board updates, and investor board reports.\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. Board narrative (planner) — lock the standard board slide order + the KPIs + the asks → gated on artifact `{{workPath}}/narrative.md` (markdown-notes)\n2. Dashboard design (designer) — dense, data-forward slide system with status colors → gated on artifact `{{workPath}}/design.md` (markdown-notes)\n3. Build the deck (developer) — implement index.html dashboard deck to the locked order + system → gated on `index.html` (slide-deck)\n\nThe gates never advance with an unmet criterion, and loop back to the\nowning phase to fix named gaps.\n",
+  "entryStepId": "narrative",
+  "triggers": [
+    "board deck",
+    "board meeting slides",
+    "board update deck",
+    "quarterly board deck",
+    "directors presentation",
+    "board report deck"
+  ],
+  "steps": [
+    {
+      "id": "narrative",
+      "name": "Board narrative",
+      "description": "lock the standard board slide order + the KPIs + the asks",
+      "prompt": "Structure the board story before any design — boards expect a consistent shape. Step 1: gather inputs: the period's headline, the KPIs that matter (with current vs target vs prior), financial summary (revenue, burn, runway), progress against the goals set last meeting, key risks, and the specific ASKS/decisions you need from the board. Step 2: lay out the canonical order: (1) title + period, (2) executive summary / TL;DR, (3) KPI dashboard, (4) financials + runway, (5) progress against last quarter's goals, (6) key risks + mitigations, (7) asks/decisions needed, (8) appendix. Step 3: for each KPI decide current value, comparison, and status (on/off track). Step 4: write the explicit asks as decisions the board must make. Step 5: write an acceptance-criteria checklist ('opens with an executive summary slide', 'KPI dashboard shows current vs target/prior with status', 'financials include runway', 'progress slide references last quarter's goals', 'a dedicated asks/decisions slide states what is needed', 'one topic per slide'). Write the slide-by-slide outline + metrics + checklist via write_task_note AND to {{workPath}}/narrative.md. No design or HTML yet.\n\nThe deliverable `{{workPath}}/narrative.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}}/narrative.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/narrative.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/narrative.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "narrative",
+        "maxAttempts": 3
+      },
+      "next": "design"
+    },
+    {
+      "id": "design",
+      "name": "Dashboard design",
+      "description": "dense, data-forward slide system with status colors",
+      "prompt": "Design a data-forward slide system suited to dense board content. Step 1: pick a restrained, serious palette with exact hex plus a clear status scale (on-track / at-risk / off-track colors used consistently). Step 2: define the grid: title + period in a header, a footer with slide number and confidentiality label, and a content zone that can hold a KPI grid or a metrics table without feeling cramped. Step 3: define slide archetypes: an executive-summary slide, a KPI dashboard (a grid of metric cards: label, big value, comparison, status dot), a financials slide (table + runway callout), a progress slide (goal -> status rows), a risks slide, and an asks slide that visually stands out. Step 4: ensure big numbers and statuses are scannable at a glance. Step 5: write a visual-criteria checklist ('consistent status colors throughout', 'KPI values are visually dominant', 'every slide has header period + footer slide number', 'asks slide visually distinct'). Write the slide-system spec + checklist via write_task_note AND to {{workPath}}/design.md. No content yet.\n\nThe deliverable `{{workPath}}/design.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": "designer",
+      "advanceWhen": {
+        "file": "{{workPath}}/design.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/design.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/design.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "design",
+        "maxAttempts": 3
+      },
+      "next": "build"
+    },
+    {
+      "id": "build",
+      "name": "Build the deck",
+      "description": "implement index.html dashboard deck to the locked order + system",
+      "prompt": "Implement the board deck as a single self-contained index.html (inline CSS + minimal JS, no external assets). Step 1: create one full-viewport section per slide in the locked board order. Step 2: apply the design system exactly — palette hex, the status color scale used consistently, header with period and footer with slide number + confidentiality label on every slide. Step 3: render the KPI dashboard as a grid of metric cards (label, dominant value, vs target/prior, status dot) and the financials slide with a table + a runway callout. Step 4: render the progress slide as goal -> status rows and make the asks/decisions slide visually distinct so directors cannot miss what is needed. Step 5: wire arrow-key navigation and visible slide numbers. Drop the approved numbers/content in verbatim; do not invent metrics. On a loop-back, fix ONLY the named gaps. Write the deck path + slide count via write_task_note.",
+      "suggestedRole": "developer",
+      "advanceWhen": {
+        "file": "index.html",
+        "minBytes": 1,
+        "sniff": "nonempty"
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "index.html",
+            "bytes": 600
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkContains",
+            "scope": "standard",
+            "inputs": {
+              "file": "index.html",
+              "pattern": "(?:slide|section|^---$)",
+              "flags": "im"
+            }
+          }
+        ],
+        "onReject": "build",
+        "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": "QA the board deck against the locked criteria. Step 1: open/render index.html and step through every slide with the arrow keys. Step 2: verify the board order (title -> exec summary -> KPI dashboard -> financials -> progress -> risks -> asks -> appendix). Step 3: check EACH acceptance criterion: opens with an executive summary, the KPI dashboard shows current vs target/prior with a consistent status color scale and dominant values, financials include runway, the progress slide references last quarter's goals, a dedicated asks/decisions slide clearly states what the board must decide and is visually distinct, every slide has the period header + slide-number footer, and there is one topic per slide. Step 4: confirm no metrics were invented beyond the outline. Write PASS/FAIL per criterion with the offending slide on any FAIL.\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: \"build\" })` 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": "build"
+    },
+    {
+      "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/bo/board-deck/versions/1.0.2/test.json

@@ -0,0 +1,103 @@
+{
+  "schemaVersion": 1,
+  "title": "Board-meeting HTML deck",
+  "objective": "Measure whether the board-deck craftbook locks the board narrative and builds a dense navigable deck with KPIs, financials, risks, asks, and appendix.",
+  "tags": [
+    "html-page"
+  ],
+  "prompt": "In the `Board Deck Eval` project, read `source/board-brief.md` and build `index.html`: a navigable Q2 board deck for Harbor & Finch with executive summary, KPI dashboard, financials, goals progress, risks, board asks, and appendix. Every slide should have a slide number and clear title. Vanilla HTML/CSS/JS only.",
+  "setup": {
+    "projectName": "Board Deck Eval",
+    "about": "Self-contained eval project for the board-deck craftbook. Board brief facts are in workspace/source/board-brief.md.",
+    "missionObjectives": "Build a navigable board deck with executive summary, KPI dashboard, financials, goals progress, risks, asks, appendix, slide numbers, and visual hierarchy.",
+    "files": [
+      {
+        "path": "source/board-brief.md",
+        "content": "# Q2 Board Brief\n\nCompany: Harbor & Finch.\nRevenue: $4.8M Q2, up 18% QoQ. Gross margin: 41%. Cash runway: 14 months.\nKPIs: first response 6 hours, refund leakage 8.9%, warehouse SLA 96%, NPS 42.\nGoals: returns automation pilot complete in West; East rollout pending Finance exception export.\nRisks: carrier scan delays in Zone 5; Monday photo-review queue exceeds SLA.\nBoard asks: approve $220K automation budget and endorse East warehouse rollout.\nAppendix: metric definitions and pilot scope.\n"
+      }
+    ],
+    "worker": {
+      "name": "Liv",
+      "role": "Developer"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "workspace/index.html is a navigable board deck with the required board sections, seeded numbers, slide numbers, and controls.",
+    "deliverables": [
+      {
+        "path": "index.html",
+        "kind": "slide-deck",
+        "minBytes": 4200,
+        "checks": [
+          {
+            "kind": "cssMinBytes",
+            "bytes": 900,
+            "file": "index.html"
+          },
+          {
+            "kind": "jsParses",
+            "file": "index.html"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "Executive summary|KPI|Financials|Goals|Risks|Board asks|Appendix",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "\\$4\\.8M|18%|41%|14 months|8\\.9%|96%|NPS 42",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "\\$220K|East warehouse rollout|Zone 5",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "index.html",
+            "pattern": "slide\\s*\\d|next|previous|keydown|arrow",
+            "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": [
+    "board narrative",
+    "KPI/financial density",
+    "asks and risks",
+    "deck navigation"
+  ]
+}