@bendyline/gilde 0.1.37 → 0.1.39

package/data/craftbook-templates/po/podcast-chaptering/versions/1.0.2/craftbook.json

@@ -0,0 +1,152 @@
+{
+  "id": "podcast-chaptering",
+  "name": "Podcast Chapters & Show Notes",
+  "description": "Turn a podcast episode into navigable chapter markers plus publish-ready show notes — timestamped chapter titles, an episode summary, key takeaways, and linked mentions. Scopes the chapter format and show-notes structure FIRST, analyzes the transcript to find real topic boundaries with timestamps, then writes the listener-facing notes in the show's voice. Separating boundary-detection (analysis) from the write-up (copy) is what yields chapters that land on actual topic shifts and notes that read like a human wrote them.\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 chapters & notes (planner) — chapter format, count, show-notes sections, voice → gated on artifact `{{workPath}}/scope.md` (markdown-notes)\n2. Find chapter boundaries (developer) — detect topic shifts, assign timestamps + titles → gated on artifact `{{workPath}}/chapters.json` (json)\n3. Write show notes (copywriter) — summary, takeaways, chapter list, links in the show voice → gated on `show-notes.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": "scope",
+  "triggers": [
+    "podcast chapters",
+    "chapter markers for my episode",
+    "write show notes",
+    "timestamp my podcast",
+    "episode summary and chapters"
+  ],
+  "steps": [
+    {
+      "id": "scope",
+      "name": "Scope chapters & notes",
+      "description": "chapter format, count, show-notes sections, voice",
+      "prompt": "Define the deliverable BEFORE analyzing. Step 1: record the source (episode title, transcript path or audio, duration). Step 2: pick the chapter marker format — the widely-supported `HH:MM:SS Title` lines (Apple/YouTube style), where the FIRST chapter must be `00:00:00`. Step 3: set chapter density (e.g. 5-12 chapters, min ~60s apart, titles ≤ ~40 chars). Step 4: define the show-notes structure: one-paragraph episode summary, 3-6 key takeaways, the timestamped chapter list, and a 'mentioned in this episode' links list. Step 5: capture the show's voice. Write an acceptance-criteria checklist (first chapter is 00:00:00, timestamps monotonic and within duration, titles concise and descriptive, notes have all four sections, every claim traces to the transcript) and decisions to `{{workPath}}/scope.md` via write_task_note AND to the produces path.\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": "planner",
+      "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": "analyze"
+    },
+    {
+      "id": "analyze",
+      "name": "Find chapter boundaries",
+      "description": "detect topic shifts, assign timestamps + titles",
+      "prompt": "Find where topics actually change. Step 1: read the transcript with its timestamps. Step 2: detect topic boundaries — points where the subject clearly shifts (new question, new segment, ad break, guest change) — not arbitrary fixed intervals. Step 3: for each boundary record the start timestamp (rounded to the second) and a short descriptive title; force the first to `00:00:00`. Step 4: enforce the scoped density/spacing/title-length rules, merging boundaries that are too close. Step 5: write the structured chapter list to `{{workPath}}/chapters.json` (each: start `HH:MM:SS`, title, a one-line topic note) via write_task_note AND the produces path so the writer has structured input. On a loop-back, adjust only the flagged boundaries.\n\nThe deliverable `{{workPath}}/chapters.json` 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}}/chapters.json",
+        "minBytes": 1,
+        "sniff": "json-valid",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/chapters.json",
+            "bytes": 2,
+            "artifact": true
+          }
+        ],
+        "onReject": "analyze",
+        "maxAttempts": 3
+      },
+      "next": "write"
+    },
+    {
+      "id": "write",
+      "name": "Write show notes",
+      "description": "summary, takeaways, chapter list, links in the show voice",
+      "prompt": "Write the publish-ready notes from the analysis. Step 1: open with a one-paragraph episode summary in the show's voice. Step 2: list 3-6 key takeaways as tight bullets. Step 3: render the chapter list as plain `HH:MM:SS Title` lines (paste-ready into a podcast host / YouTube description) directly from `{{workPath}}/chapters.json`, first line `00:00:00`. Step 4: add a 'Mentioned in this episode' section with the people/links/resources referenced. Step 5: assemble all of it into `show-notes.md` with clear section headers. Do not invent facts not in the transcript. write_task_note the path and chapter count. On a loop-back, fix only the flagged sections.\n\nBefore working, open the artifact input `{{workPath}}/chapters.json` with `read_artifact`; do not look for it in the workspace.",
+      "suggestedRole": "copywriter",
+      "consumes": [
+        {
+          "file": "{{workPath}}/chapters.json",
+          "artifact": true
+        }
+      ],
+      "advanceWhen": {
+        "file": "show-notes.md",
+        "minBytes": 1,
+        "sniff": "nonempty"
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "show-notes.md",
+            "bytes": 500
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkContains",
+            "scope": "standard",
+            "inputs": {
+              "file": "show-notes.md",
+              "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+              "flags": "i"
+            }
+          }
+        ],
+        "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": "Check the chapters and notes. Step 1: confirm the chapter list's FIRST line is `00:00:00` and all timestamps are monotonic, spaced per scope, and within the episode duration (PASS/FAIL). Step 2: spot-check 3 chapter timestamps against the transcript — does each land on a real topic shift? Step 3: confirm show-notes.md has all four sections (summary, takeaways, chapters, mentions). Step 4: confirm claims trace to the transcript (nothing invented) and the voice matches the show. Step 5: confirm titles are concise and descriptive. Write PASS/FAIL per criterion.\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": "show-notes.md"
+        }
+      ],
+      "next": "write"
+    },
+    {
+      "id": "finish",
+      "name": "Finish",
+      "description": "All acceptance criteria met. Stamp a short summary and report DONE.",
+      "prompt": "Every acceptance criterion passed. Write a one-paragraph DONE summary to task notes via `write_task_note`: what was built, the deliverable path(s), and a one-line confirmation that each criterion is met. Then report DONE.",
+      "suggestedRole": "developer",
+      "terminal": true
+    }
+  ],
+  "version": "1.0.2",
+  "releasedAt": "2026-08-21T00:00:00Z",
+  "paramSchema": {
+    "type": "object",
+    "properties": {
+      "workPath": {
+        "type": "string",
+        "title": "Working folder",
+        "description": "Per-task working folder in the artifacts drawer. Defaults to this task's own folder so runs never collide; override with a stable name when you deliberately want runs to share files.",
+        "default": "{{task.dir}}"
+      }
+    }
+  },
+  "minGezelVersion": "1.26233"
+}

package/data/craftbook-templates/po/podcast-chaptering/versions/1.0.2/test.json

@@ -0,0 +1,106 @@
+{
+  "schemaVersion": 1,
+  "title": "Podcast Chapters & Show Notes smoke eval",
+  "objective": "Self-contained smoke eval for the Podcast Chapters & Show Notes craftbook using the media generic harness.",
+  "tags": [
+    "media"
+  ],
+  "prompt": "Our media files are all listd in source/media-index.csv. Can you put together the production plan for them as media-plan.json?",
+  "setup": {
+    "projectName": "Podcast Chapters & Show Notes Eval",
+    "about": "Self-contained eval project for podcast-chaptering. Seeded inputs are under workspace/source or workspace/fixtures; final deliverable is workspace/media-plan.json.",
+    "missionObjectives": "Use the Podcast Chapters & Show Notes craftbook/template, read the seeded local fixtures, and write media-plan.json without network calls, real credentials, or live services.",
+    "files": [
+      {
+        "path": "source/brief.md",
+        "content": "# Podcast Chapters & Show Notes 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: podcast-chaptering - Podcast Chapters & Show Notes.\n"
+      },
+      {
+        "path": "source/media-index.csv",
+        "content": "filename,type,duration_or_size,notes\npilot-interview.wav,audio,00:02:15,mentions response time and status silence\nreturns-desk.png,image,1600x900,shows intake form and photo evidence step\nlaunch-clip.mp4,video,00:01:05,shows barcode exception handling\n"
+      }
+    ],
+    "worker": {
+      "name": "Jules",
+      "role": "Producer"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "media-plan.json is a valid JSON media-production manifest grounded in seeded fixture metadata.",
+    "deliverables": [
+      {
+        "path": "media-plan.json",
+        "kind": "json",
+        "minBytes": 700,
+        "checks": [
+          {
+            "kind": "sniff",
+            "file": "media-plan.json",
+            "sniff": "json-valid"
+          },
+          {
+            "kind": "contains",
+            "file": "media-plan.json",
+            "pattern": "pilot-interview\\.wav|returns-desk\\.png|launch-clip\\.mp4",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "media-plan.json",
+            "pattern": "outputs|assets|steps|cues|verification",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "media-plan.json",
+            "pattern": "accessibility|alt|caption|transcript|metadata",
+            "flags": "i"
+          }
+        ]
+      }
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "media-plan.json",
+      "kind": "json"
+    },
+    "axes": [
+      {
+        "name": "completeness",
+        "description": "The manifest covers sources, outputs, and processing steps end to end."
+      },
+      {
+        "name": "grounding",
+        "description": "Only seeded media fixtures are referenced, with faithful metadata."
+      },
+      {
+        "name": "accessibility",
+        "description": "Alt text, captions, or transcripts are planned wherever the class requires."
+      },
+      {
+        "name": "precision",
+        "description": "Timestamps, crops, and cues are specific enough to execute."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "media fixture handling",
+    "structured manifest",
+    "verification metadata"
+  ],
+  "extensions": {
+    "legacySimulators": [
+      {
+        "id": "podcast-chaptering-media-fixtures",
+        "kind": "data-source",
+        "status": "implemented",
+        "description": "Seeded local media index replacing real audio, video, image generation, and external media metadata services."
+      }
+    ]
+  }
+}

package/data/craftbook-templates/po/portfolio-site/versions/1.0.2/craftbook.json

@@ -0,0 +1,163 @@
+{
+  "id": "portfolio-site",
+  "name": "Portfolio Site",
+  "description": "Build a personal or professional portfolio site that showcases projects, an about/bio, and contact links in a single polished page. Locks the content inventory — which projects, what bio, which links — first, then a clean gallery-style visual design, then builds responsive HTML with a project grid and real CSS. Scoping the content before designing prevents a small model from inventing fake projects or building a layout the real content does not fit; the portfolio's job is to present a finite set of real work cleanly.\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. Content scope (planner) — inventory projects, bio, links, sections → gated on artifact `{{workPath}}/content-scope.md` (markdown-notes)\n2. Visual design (visual-designer) — gallery layout, project card, palette, type → gated on artifact `{{workPath}}/visual-design.md` (markdown-notes)\n3. Build (developer) — implement index.html with the project grid → gated on `index.html` (html-marketing-site)\n\nThe gates never advance with an unmet criterion, and loop back to the\nowning phase to fix named gaps.\n",
+  "entryStepId": "content-scope",
+  "triggers": [
+    "portfolio site",
+    "personal website",
+    "showcase my work",
+    "developer portfolio",
+    "designer portfolio",
+    "build a portfolio"
+  ],
+  "steps": [
+    {
+      "id": "content-scope",
+      "name": "Content scope",
+      "description": "inventory projects, bio, links, sections",
+      "prompt": "Inventory the real content. (1) List the projects to feature (title, one-line description, role, link/repo, and whether an image exists) — use only what the user provided; if details are missing, leave placeholders the user must fill, do NOT fabricate. (2) Draft the bio focus (who, what they do, what they want). (3) List contact/social links to include. (4) Order the sections: hero/intro, project grid, about, contact. (5) Write an acceptance-criteria checklist of 5-8 items, e.g. 'every listed project has a card', 'the hero states name and role', 'an about section is present', 'contact links are present and clickable', 'responsive at 375px and 1280px', 'no fabricated projects'. Write to {{workPath}}/content-scope.md and via write_task_note. No design or HTML yet.\n\nThe deliverable `{{workPath}}/content-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}}/content-scope.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/content-scope.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/content-scope.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "content-scope",
+        "maxAttempts": 3
+      },
+      "next": "visual-design"
+    },
+    {
+      "id": "visual-design",
+      "name": "Visual design",
+      "description": "gallery layout, project card, palette, type",
+      "prompt": "Design the showcase look. (1) Pick a restrained palette (hex) and type pairing that lets the work shine, not the chrome. (2) Design the project card (image/placeholder, title, description, link) and how cards lay out in a responsive grid. (3) Lay out the hero/intro and the about/contact sections. (4) Specify hover/focus states for cards and links and the overall spacing rhythm. (5) Self-contained CSS, no external assets beyond any provided images. Write the visual spec to {{workPath}}/visual-design.md and via write_task_note. No HTML yet.\n\nThe deliverable `{{workPath}}/visual-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": "visual-designer",
+      "advanceWhen": {
+        "file": "{{workPath}}/visual-design.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/visual-design.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/visual-design.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "visual-design",
+        "maxAttempts": 3
+      },
+      "next": "build"
+    },
+    {
+      "id": "build",
+      "name": "Build",
+      "description": "implement index.html with the project grid",
+      "prompt": "Implement index.html with inline CSS to the visual spec, using only the scoped content. (1) Palette as CSS variables and the chosen type. (2) Hero with name and role. (3) A responsive CSS-grid project gallery, one card per scoped project with its real (or placeholder) details and a clickable link. (4) About and contact sections with clickable links. (5) Responsive at 375px and 1280px; vanilla HTML/CSS, no CDN. Keep placeholders explicit where the user must supply content — never invent projects. On a loop-back, fix only the named gaps without regressing passing criteria. write_task_note the path and passing criteria.",
+      "suggestedRole": "developer",
+      "advanceWhen": {
+        "file": "index.html",
+        "minBytes": 1,
+        "sniff": "html-complete"
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "index.html",
+            "bytes": 1200
+          },
+          {
+            "kind": "htmlLint",
+            "file": "index.html"
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkHtmlComplete",
+            "scope": "standard",
+            "inputs": {
+              "file": "index.html"
+            }
+          },
+          {
+            "name": "checkCssMinBytes",
+            "scope": "standard",
+            "inputs": {
+              "file": "index.html",
+              "minBytes": 600
+            }
+          }
+        ],
+        "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": "Evaluate the portfolio. Open/render index.html and resize to 375px and 1280px. Check EACH criterion from {{workPath}}/content-scope.md and write PASS/FAIL with a reason: does every scoped project have a card, does the hero state name and role, is there an about section, are contact links present and clickable, is it responsive, and crucially are there ZERO fabricated projects? Any invented project or missing scoped project is a FAIL — loop back to build.\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/po/portfolio-site/versions/1.0.2/test.json

@@ -0,0 +1,100 @@
+{
+  "schemaVersion": 1,
+  "title": "Portfolio Site smoke eval",
+  "objective": "Self-contained smoke eval for the Portfolio Site 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": "Portfolio Site Eval",
+    "about": "Self-contained eval project for portfolio-site. Seeded inputs are under workspace/source or workspace/fixtures; final deliverable is workspace/index.html.",
+    "missionObjectives": "Use the Portfolio Site 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": "# Portfolio Site 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: portfolio-site - Portfolio Site.\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/po/powerpoint-deck/versions/1.7.3/craftbook.json

@@ -1,7 +1,7 @@
 {
   "id": "powerpoint-deck",
   "name": "PowerPoint from Content",
-  "description": "Turn supplied content or a factual topic into a real, editable, source-grounded PowerPoint. A researcher first reads the exact named source or retrieves authoritative web sources and records a cited fact ledger. A planner locks a numbered slide outline. A copywriter writes a task-specific Markdown deck with exactly one H1 per outline slide, while deterministic gates keep the slide count, titles, and order aligned. A reviewer checks factual traceability before a designer converts the Markdown to PPTX, previews every slide, and saves the requested PPTX in artifacts.\n\nEvery run gets its own workPath by default, so source packets, outlines, Markdown, and reviews cannot collide with an earlier PowerPoint task. Artifact working files must be read and written with the artifact tools; the Markdown deck source and user-facing PowerPoint remain in the workspace.",
+  "description": "Turn supplied content or a factual topic into a real, editable PowerPoint. Supplied sources remain authoritative, and internet research is strongly recommended for factual grounding; topic-only runs may proceed without it when access is unavailable or the deck is explicitly exploratory, provided unsupported facts are not invented. A researcher records the source boundary and fact ledger before a planner locks a numbered slide outline. A copywriter writes a task-specific Markdown deck with exactly one H1 per outline slide, while deterministic gates keep the slide count, titles, and order aligned. A reviewer checks factual traceability before a designer converts the Markdown to PPTX, previews every slide, and saves the requested PPTX in artifacts.\n\nEvery run gets its own workPath by default, so source packets, outlines, Markdown, and reviews cannot collide with an earlier PowerPoint task. Artifact working files must be read and written with the artifact tools; the Markdown deck source and user-facing PowerPoint remain in the workspace.",
   "entryStepId": "research",
   "triggers": [
     "make a powerpoint",
@@ -99,8 +99,8 @@
     {
       "id": "research",
       "name": "Acquire and verify sources",
-      "description": "Read the exact supplied source or retrieve authoritative sources before outlining.",
-      "prompt": "Build the source packet before planning any slides. This run's working namespace is `{{workPath}}`. Files outside it—including an earlier `notes/outline.md`, root `deck.md`, or another PowerPoint task directory—are not inputs and must not be read or reused unless `{{sourcePath}}` names one exactly. Exact source path: `{{sourcePath}}`. Topic: `{{topic}}`. Inline source brief: `{{content}}`. Audience: `{{audience}}`. 0. If source path, topic, and inline content are all empty, call `ask_user_question` for the missing subject or exact source and stop. Do not call `read_task_notes`, inspect the workspace, or reuse an earlier deck while the source boundary is missing. 1. When `{{sourcePath}}` is non-empty, your FIRST source action is `read_file` on that exact path. Never replace a named source with a similarly named file. 2. Otherwise treat non-empty `{{content}}` as supplied source material and `{{topic}}` as the research question, but still perform real external source acquisition for factual grounding: use a registered search tool such as `web_search` or `wikipedia_search`, then use `fetch_url` or browser automation to read the useful result. Prefer primary sources, official statistics, standards bodies, peer-reviewed work, or reputable institutional summaries. Do not treat model memory, the task prompt, an earlier deck, or an invented URL as research. 3. Write `{{workPath}}/sources.md` with: the invocation inputs and their boundary; a source table with title, resolvable URL or exact file path, authority/date, and intended use; a fact ledger where every number, quotation, and externally checkable claim carries `(source: URL-or-path)`; conflicts and uncertainties; and facts that must not be invented. Do not silently overwrite supplied facts with external claims—record conflicts explicitly. 4. Include at least one resolvable citation and call `write_task_note` with the sources used and any uncertainty. Do not draft the outline or deck yet.\n\nStore `{{workPath}}/sources.md` in the artifacts drawer with `write_artifact`, then verify it with `read_artifact`; do not write this working file to the workspace.",
+      "description": "Read the exact supplied source and strongly prefer authoritative research before outlining, while allowing an honest no-research path.",
+      "prompt": "Build the source packet before planning any slides. This run's working namespace is `{{workPath}}`. Files outside it—including an earlier `notes/outline.md`, root `deck.md`, or another PowerPoint task directory—are not inputs and must not be read or reused unless `{{sourcePath}}` names one exactly. Exact source path: `{{sourcePath}}`. Topic: `{{topic}}`. Inline source brief: `{{content}}`. Audience: `{{audience}}`. 0. If source path, topic, and inline content are all empty, call `ask_user_question` for the missing subject or exact source and stop. Do not call `read_task_notes`, inspect the workspace, or reuse an earlier deck while the source boundary is missing. 1. When `{{sourcePath}}` is non-empty, your FIRST source action is `read_file` on that exact path. Never replace a named source with a similarly named file, and never skip a supplied source. 2. Otherwise treat non-empty `{{content}}` as supplied source material and `{{topic}}` as the research question. Strongly prefer real external source acquisition for factual grounding: use `web_search` when registered, otherwise `wikipedia_search`, then use `fetch_url` or browser automation to read the useful result. Prefer primary sources, official statistics, standards bodies, peer-reviewed work, or reputable institutional summaries. You may proceed without external research when external tools are unavailable or when the topic is explicitly exploratory, generative, or proposal-oriented. If you skip it, say so plainly, record why, frame unsourced content as original concepts or reasoning rather than established fact, and do not invent a source, URL, research result, or tool-verification workaround. Do not treat model memory, the task prompt, an earlier deck, or an invented URL as research. 3. Write `{{workPath}}/sources.md` with: the invocation inputs and their boundary; research status (`completed` or `skipped`, with reason); a source table with title, resolvable URL or exact file path, authority/date, and intended use (the table may be empty when research was honestly skipped); a fact ledger where every number, quotation, and externally checkable claim carries `(source: URL-or-path)`; a separate ledger of original proposals or reasoning that makes no factual-source claim; conflicts and uncertainties; and facts that must not be invented. Do not silently overwrite supplied facts with external claims—record conflicts explicitly. 4. When any source was used, include at least one resolvable citation. When none was used, include no citation rather than fabricating one. Call `write_task_note` with the sources used—or that research was skipped—and any uncertainty. Do not draft the outline or deck yet.\n\nStore `{{workPath}}/sources.md` in the artifacts drawer with `write_artifact`, then verify it with `read_artifact`; do not write this working file to the workspace.",
       "suggestedRole": "researcher",
       "advanceWhen": {
         "file": "{{workPath}}/sources.md",
@@ -120,7 +120,7 @@
           {
             "kind": "citationsResolve",
             "file": "{{workPath}}/sources.md",
-            "minCitations": 1,
+            "minCitations": 0,
             "artifact": true
           },
           {
@@ -133,7 +133,8 @@
               "browser_navigate",
               "run_playwright_script"
             ],
-            "minSuccessful": 1
+            "minSuccessful": 1,
+            "externalOptional": true
           }
         ],
         "onReject": "research",
@@ -145,7 +146,7 @@
       "id": "outline",
       "name": "Lock the slide outline",
       "description": "Lock the audience, purpose, sequence, evidence, and one message per slide.",
-      "prompt": "Read `{{workPath}}/sources.md` and plan the deck from its verified fact ledger. If `{{sourcePath}}` is non-empty, also read that exact source. Do not borrow facts, titles, or slide content from an earlier outline or deck outside `{{workPath}}`; a requested similarity is a structural constraint, not permission to mix source packets. 1. State the audience, occasion, and ONE takeaway. 2. Use exactly the slide count the user asked for; only when they gave none, draft 8-15 slides. Format every slide heading exactly as `## Slide N — Title`, numbered consecutively from 1. Under each heading write `Message:`, `Evidence:`, and `Visual:`. Every evidence line must point to a citation already recorded in `{{workPath}}/sources.md`; mark unsupported ideas `UNVERIFIED` rather than filling gaps from memory. 3. Open with context, make the middle a coherent argument, and close with the ask or next step. Include a Sources slide in this locked outline when the audience needs visible references. 4. Write acceptance criteria: every numbered outline slide must become exactly one H1 in {{workPath}}/deck.md, with the same title and order; one idea per slide; no wall of text; claims remain traceable; visual preview passes; an artifact copy is retained; and the exact workspace file `{{outputPath}}` is delivered. Write only `{{workPath}}/outline.md`, then call `write_task_note` with the exact slide count, takeaway, and requested output path. Do not write slide content or presentation code.\n\nOpen `{{workPath}}/sources.md` from the artifacts drawer with `read_artifact` before working. Store `{{workPath}}/outline.md` in the artifacts drawer with `write_artifact`, then verify it with `read_artifact`; do not write this working file to the workspace.",
+      "prompt": "Read `{{workPath}}/sources.md` and plan the deck from its verified fact ledger and original-proposal ledger. If `{{sourcePath}}` is non-empty, also read that exact source. Do not borrow facts, titles, or slide content from an earlier outline or deck outside `{{workPath}}`; a requested similarity is a structural constraint, not permission to mix source packets. 1. State the audience, occasion, and ONE takeaway. 2. Use exactly the slide count the user asked for; only when they gave none, draft 8-15 slides. Format every slide heading exactly as `## Slide N — Title`, numbered consecutively from 1. Under each heading write `Message:`, `Evidence:`, and `Visual:`. Every sourced factual evidence line must point to a citation already recorded in `{{workPath}}/sources.md`. When research was skipped, original concepts and reasoned recommendations may be labeled `PROPOSAL` and need no citation; mark any unsupported externally checkable claim `UNVERIFIED` rather than filling gaps from memory. 3. Open with context, make the middle a coherent argument, and close with the ask or next step. Include a Sources slide in this locked outline when the audience needs visible references and sources actually exist. 4. Write acceptance criteria: every numbered outline slide must become exactly one H1 in {{workPath}}/deck.md, with the same title and order; one idea per slide; no wall of text; factual claims remain traceable; proposals remain clearly framed as proposals; visual preview passes; an artifact copy is retained; and the exact workspace file `{{outputPath}}` is delivered. Write only `{{workPath}}/outline.md`, then call `write_task_note` with the exact slide count, takeaway, and requested output path. Do not write slide content or presentation code.\n\nOpen `{{workPath}}/sources.md` from the artifacts drawer with `read_artifact` before working. Store `{{workPath}}/outline.md` in the artifacts drawer with `write_artifact`, then verify it with `read_artifact`; do not write this working file to the workspace.",
       "suggestedRole": "planner",
       "advanceWhen": {
         "file": "{{workPath}}/outline.md",
@@ -173,7 +174,7 @@
           {
             "kind": "citationsResolve",
             "file": "{{workPath}}/outline.md",
-            "minCitations": 1,
+            "minCitations": 0,
             "artifact": true
           }
         ],
@@ -222,7 +223,7 @@
           {
             "kind": "citationsResolve",
             "file": "{{workPath}}/deck.md",
-            "minCitations": 1
+            "minCitations": 0
           }
         ],
         "onReject": "write",