@bendyline/gilde 0.1.39 → 0.1.41

package/data/craftbook-templates/re/refactor-module/versions/2.0.0/test.json

@@ -0,0 +1,242 @@
+{
+  "schemaVersion": 1,
+  "title": "Behavior-preserving extraction of duplicated logic",
+  "objective": "The craftbook drives a disciplined refactor of copy-pasted name-normalization logic: a green baseline recorded before anything moves, reversible stages with the suite green throughout, a re-measured duplication target, and an enforced review — verified mechanically via a reference matrix across both call sites, a duplication count across src/, and a green-suite run.",
+  "tags": [
+    "code",
+    "refactor",
+    "tactical-fleet"
+  ],
+  "prompt": "Use the Behavior-Preserving Refactor craftbook for this: the same name-normalization logic is copy-pasted in src/format-name.js and src/report.js, and the copies have drifted once already. The brief is in source/refactor-brief.md. Extract the shared logic into a new src/normalize-name.js used by both files, without changing behavior.",
+  "setup": {
+    "projectName": "Crewlog",
+    "about": "A tiny crew-hours logging module. Tests run with `npm run test` (node --test).",
+    "files": [
+      {
+        "path": "package.json",
+        "content": "{\n  \"name\": \"crewlog\",\n  \"private\": true,\n  \"version\": \"1.0.0\",\n  \"type\": \"module\",\n  \"scripts\": {\n    \"test\": \"node --test\"\n  }\n}\n"
+      },
+      {
+        "path": "src/format-name.js",
+        "content": "export function formatName(raw) {\n  const words = String(raw).trim().split(/\\s+/);\n  const fixed = [];\n  for (const word of words) {\n    fixed.push(word.charAt(0).toUpperCase() + word.slice(1).toLowerCase());\n  }\n  return fixed.join(' ');\n}\n"
+      },
+      {
+        "path": "src/report.js",
+        "content": "function titleCaseName(raw) {\n  const words = String(raw).trim().split(/\\s+/);\n  const fixed = [];\n  for (const word of words) {\n    fixed.push(word.charAt(0).toUpperCase() + word.slice(1).toLowerCase());\n  }\n  return fixed.join(' ');\n}\n\nexport function reportLine(entry) {\n  return titleCaseName(entry.name) + ': ' + entry.hours + 'h';\n}\n\nexport function reportSummary(entries) {\n  return entries.map((entry) => reportLine(entry)).join('\\n');\n}\n"
+      },
+      {
+        "path": "tests/format.test.mjs",
+        "content": "import assert from 'node:assert/strict';\nimport { test } from 'node:test';\nimport { formatName } from '../src/format-name.js';\n\ntest('formats a messy name', () => {\n  assert.equal(formatName('  ada   lovelace '), 'Ada Lovelace');\n});\n"
+      },
+      {
+        "path": "tests/report.test.mjs",
+        "content": "import assert from 'node:assert/strict';\nimport { test } from 'node:test';\nimport { reportLine } from '../src/report.js';\n\ntest('report line normalizes the name', () => {\n  assert.equal(reportLine({ name: 'grace hopper', hours: 2 }), 'Grace Hopper: 2h');\n});\n"
+      },
+      {
+        "path": "source/refactor-brief.md",
+        "content": "# Refactor brief\n\nThe same name-normalization logic (trim, collapse inner whitespace, title-case each word) is copy-pasted in two places:\n\n- src/format-name.js - `formatName`, used by the profile screens\n- src/report.js - a private `titleCaseName` helper feeding `reportLine`\n\nThe copies drifted once before and had to be manually re-synced. Extract the shared logic into a NEW `src/normalize-name.js` module used by both files, WITHOUT changing behavior. Every output must stay byte-for-byte identical, including edge cases: extra inner spaces, ALL-CAPS input, single letters, and the empty string. Public exports must survive unchanged: `formatName`, `reportLine`, and `reportSummary` keep their names and signatures.\n"
+      },
+      {
+        "path": "tests/verify-refactor.mjs",
+        "content": "import assert from 'node:assert/strict';\nimport { execFileSync } from 'node:child_process';\nimport { readFileSync, readdirSync } from 'node:fs';\nimport { join } from 'node:path';\nimport { formatName } from '../src/format-name.js';\nimport { reportLine, reportSummary } from '../src/report.js';\n\n// 1) Behavior: a reference matrix across BOTH call sites, quirks included.\nconst matrix = [\n  ['  ada   lovelace ', 'Ada Lovelace'],\n  ['GRACE HOPPER', 'Grace Hopper'],\n  ['katherine  g.  johnson', 'Katherine G. Johnson'],\n  ['a', 'A'],\n  ['', ''],\n];\nfor (const [input, expected] of matrix) {\n  assert.equal(formatName(input), expected, 'formatName(' + JSON.stringify(input) + ')');\n}\nassert.equal(reportLine({ name: '  grace   HOPPER ', hours: 4 }), 'Grace Hopper: 4h');\nassert.equal(reportLine({ name: 'ada lovelace', hours: 2 }), 'Ada Lovelace: 2h');\nassert.equal(\n  reportSummary([\n    { name: 'ADA   LOVELACE', hours: 1 },\n    { name: ' mary jackson ', hours: 3 },\n  ]),\n  'Ada Lovelace: 1h\\nMary Jackson: 3h',\n);\nconsole.log('REFACTOR_MODULE_ORACLE behavior ok');\n\n// 2) Deduplication: the title-case implementation exists at most once\n// across src/ - two surviving copies mean the extraction did not happen.\nconst root = process.cwd();\nfunction countMarker(relDir) {\n  let count = 0;\n  for (const entry of readdirSync(join(root, relDir), { withFileTypes: true })) {\n    const relPath = relDir + '/' + entry.name;\n    if (entry.isDirectory()) {\n      count += countMarker(relPath);\n    } else if (/\\.(js|mjs|cjs|ts)$/.test(entry.name)) {\n      const text = readFileSync(join(root, relPath), 'utf8');\n      count += (text.match(/toUpperCase/g) ?? []).length;\n    }\n  }\n  return count;\n}\nconst occurrences = countMarker('src');\nassert.ok(\n  occurrences <= 1,\n  'the title-case implementation appears ' + occurrences + ' times across src/ - the duplication was not extracted to one shared place',\n);\nconsole.log('REFACTOR_MODULE_ORACLE dedupe ok');\n\n// 3) The whole suite is green on the refactored tree (real run).\nfunction runSuite(cwd) {\n  try {\n    execFileSync(process.execPath, ['--test'], { cwd, stdio: 'pipe', timeout: 30000 });\n    return 0;\n  } catch (err) {\n    return typeof err.status === 'number' ? err.status : 1;\n  }\n}\nassert.equal(runSuite(root), 0, 'the suite must be green on the refactored tree');\nconsole.log('REFACTOR_MODULE_ORACLE suite green');\nconsole.log('REFACTOR_MODULE_ORACLE done');\n",
+        "surface": "harness"
+      }
+    ],
+    "craftbookParams": {
+      "workPath": "tasks/eval",
+      "scope": "Extract the duplicated name normalization in src/format-name.js and src/report.js into a shared src/normalize-name.js used by both, without changing behavior; see source/refactor-brief.md."
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "The duplicated normalization lives in exactly one place, both call sites behave byte-for-byte as before (reference matrix), the suite is green on a real run, and every fleet artifact exists with honest citations.",
+    "deliverables": [
+      {
+        "path": "tasks/eval/baseline.md",
+        "kind": "markdown-notes",
+        "artifact": true,
+        "minBytes": 500,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "tasks/eval/baseline.md",
+            "pattern": "^##\\s+Current behavior[\\s\\S]*^##\\s+Measurements[\\s\\S]*^##\\s+Guardrail",
+            "flags": "im",
+            "label": "baseline sections in order"
+          },
+          {
+            "kind": "citationsResolve",
+            "file": "tasks/eval/baseline.md",
+            "minCitations": 2,
+            "artifact": true
+          }
+        ]
+      },
+      {
+        "path": "tasks/eval/plan.md",
+        "kind": "markdown-notes",
+        "artifact": true,
+        "minBytes": 500,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "tasks/eval/plan.md",
+            "pattern": "^##\\s+Target[\\s\\S]*^##\\s+Stages[\\s\\S]*^##\\s+Acceptance criteria",
+            "flags": "im",
+            "label": "plan sections in order"
+          },
+          {
+            "kind": "citationsResolve",
+            "file": "tasks/eval/plan.md",
+            "minCitations": 2,
+            "artifact": true
+          }
+        ]
+      },
+      {
+        "path": "tasks/eval/change-notes.md",
+        "kind": "markdown-notes",
+        "artifact": true,
+        "minBytes": 600,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "tasks/eval/change-notes.md",
+            "pattern": "^##\\s+Stages executed[\\s\\S]*^##\\s+Files touched[\\s\\S]*^##\\s+Deviations from plan",
+            "flags": "im",
+            "label": "change-notes sections in order"
+          },
+          {
+            "kind": "citationsResolve",
+            "file": "tasks/eval/change-notes.md",
+            "minCitations": 2,
+            "artifact": true
+          }
+        ]
+      },
+      {
+        "path": "tasks/eval/verification.md",
+        "kind": "markdown-notes",
+        "artifact": true,
+        "minBytes": 400,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "tasks/eval/verification.md",
+            "pattern": "^##\\s+Before\\s*/\\s*after[\\s\\S]*^##\\s+Suite[\\s\\S]*^##\\s+Result",
+            "flags": "im",
+            "label": "verification sections in order"
+          },
+          {
+            "kind": "contains",
+            "file": "tasks/eval/verification.md",
+            "pattern": "^##\\s+Before\\s*/\\s*after[\\s\\S]*\\d",
+            "flags": "im",
+            "label": "before/after figures quoted"
+          }
+        ]
+      },
+      {
+        "path": "tasks/eval/review.md",
+        "kind": "markdown-report",
+        "artifact": true,
+        "minBytes": 400,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "tasks/eval/review.md",
+            "pattern": "Verdict:\\s*(?:PASS|REVISE)",
+            "flags": "i",
+            "label": "explicit reviewer verdict"
+          }
+        ]
+      }
+    ],
+    "checks": [
+      {
+        "kind": "notContains",
+        "file": "src/format-name.js",
+        "pattern": "toUpperCase",
+        "label": "format-name.js no longer carries its own title-case copy"
+      },
+      {
+        "kind": "notContains",
+        "file": "src/report.js",
+        "pattern": "toUpperCase",
+        "label": "report.js no longer carries its own title-case copy"
+      },
+      {
+        "kind": "sourceParses",
+        "file": "src/format-name.js"
+      },
+      {
+        "kind": "sourceParses",
+        "file": "src/report.js"
+      },
+      {
+        "kind": "nodeScriptPasses",
+        "script": "tests/verify-refactor.mjs",
+        "timeoutMs": 60000,
+        "requiredOutput": [
+          {
+            "pattern": "REFACTOR_MODULE_ORACLE behavior ok",
+            "label": "both call sites match the reference matrix"
+          },
+          {
+            "pattern": "REFACTOR_MODULE_ORACLE dedupe ok",
+            "label": "the implementation exists at most once across src/"
+          },
+          {
+            "pattern": "REFACTOR_MODULE_ORACLE suite green",
+            "label": "suite green on the refactored tree"
+          }
+        ]
+      }
+    ],
+    "taskNotes": {
+      "minBytes": 120,
+      "checks": [
+        {
+          "kind": "contains",
+          "file": "task-notes.md",
+          "pattern": "\\bDONE\\b[\\s\\S]*(npm\\s+run\\s+test|node\\s+--test)",
+          "flags": "i",
+          "label": "DONE note names the real test command"
+        }
+      ]
+    },
+    "taskGraph": {
+      "requireCraftbookTask": true,
+      "requireTerminalStep": true
+    },
+    "unchangedFixtures": [
+      "source/refactor-brief.md",
+      "package.json"
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "tasks/eval/change-notes.md",
+      "kind": "markdown"
+    },
+    "axes": [
+      {
+        "name": "Guardrail-first discipline",
+        "description": "The suite was genuinely run and recorded green before any restructuring, characterization coverage was considered honestly, and the baseline counted the duplication with a stated method."
+      },
+      {
+        "name": "Reversible-stage execution",
+        "description": "The restructure happened in small planned stages at the real sites, with the suite kept green between stages and deviations recorded honestly rather than papered over."
+      },
+      {
+        "name": "Genuine deduplication",
+        "description": "One shared implementation remains, both call sites use it, public exports survived unchanged, and no behavior change was smuggled in under cover of the refactor."
+      }
+    ]
+  },
+  "qualityFocus": [
+    "green-before/green-after receipts",
+    "same-method re-measurement",
+    "enforced review loop"
+  ]
+}

package/data/craftbook-templates/re/regex-builder/manifest.json

@@ -3,6 +3,7 @@
   "kind": "craftbook-template",
   "id": "regex-builder",
   "role": "general",
+  "category": "code-data",
   "name": "Build + Test a Tricky Regex",
   "description": "Build a regular expression that matches exactly the right strings and rejects the wrong ones — specified with concrete should-match and should-NOT-match cases before a single character is written. Specs the cases FIRST — the positive examples, the tricky negative examples, edge cases, and the anchoring/flags — then builds the regex incrementally against those cases, then tests the full matrix including catastrophic-backtracking safety. Covers character classes, quantifiers, anchors, groups/captures, lookahead/behind, escaping, flags, and ReDoS-safe patterns validated by a should-match / should-not-match table.",
   "tags": [

package/data/craftbook-templates/re/release-announcement/manifest.json

@@ -3,6 +3,7 @@
   "kind": "craftbook-template",
   "id": "release-announcement",
   "role": "general",
+  "category": "marketing",
   "name": "Release Announcement",
   "description": "Write a public-facing release announcement that turns a shipped version into a compelling story for users — a blog-post / email-style piece that leads with the headline value, shows the marquee features with concrete before/after benefits, links to the full release notes and upgrade path, and ends with a clear call to action. A planner picks the 2-4 things actually worth announcing and the audience/channel, a copywriter writes the benefit-led announcement in the brand voice, and a reviewer checks it is accurate, on-message, and that every claim traces to a real shipped change. Use this for product launch posts, 'what's new' announcements, version launch emails, and changelog-to-marketing translations.",
   "tags": [

package/data/craftbook-templates/re/release-artifact-sanity-check/manifest.json

@@ -3,6 +3,7 @@
   "kind": "craftbook-template",
   "id": "release-artifact-sanity-check",
   "role": "maintenance-review",
+  "category": "code-release",
   "name": "Release Artifact Sanity Check",
   "description": "Sanity-check a folder of release artifacts without executing them: inventory every package, verify hashes and signatures against trusted evidence, safely inspect every file inside archives and application bundles, scan for suspicious or secret-bearing content, and produce an evidence-backed ship or no-ship report. Use this for release artifact verification, installer signing checks, package integrity audits, supply-chain inspection, or a pre-publish binary review.",
   "tags": [

package/data/craftbook-templates/re/release-notes/manifest.json

@@ -3,6 +3,7 @@
   "kind": "craftbook-template",
   "id": "release-notes",
   "role": "general",
+  "category": "code-release",
   "name": "Release Notes",
   "description": "Turn a raw set of merged commits, pull requests, and ticket IDs into polished, human-readable release notes for a versioned release — grouped by theme (Features, Improvements, Bug Fixes, Breaking Changes), written for users rather than engineers, with upgrade/migration callouts surfaced first. The ordering is deliberate: a developer gathers and de-noises the change inventory before any prose is written, a copywriter turns that inventory into benefit-led notes in a consistent voice, and a reviewer verifies every shipped change is represented and nothing internal-only or unreleased leaked in. Use this for GitHub Releases, product announcement posts, or the user-facing notes that ship alongside a tagged version.",
   "tags": [
@@ -11,7 +12,11 @@
     "release-notes",
     "changelog",
     "announcement",
-    "versioning"
+    "versioning",
+    "semver",
+    "keep-a-changelog",
+    "writing",
+    "developer"
   ],
   "maintainer": {
     "name": "Gezel"

package/data/craftbook-templates/re/release-notes/versions/1.0.3/craftbook.json

@@ -0,0 +1,220 @@
+{
+  "id": "release-notes",
+  "name": "Release Notes",
+  "description": "Turn a raw set of merged commits, pull requests, and ticket IDs into polished, human-readable release notes for a versioned release — grouped by theme (Features, Improvements, Bug Fixes, Breaking Changes), written for users rather than engineers, with upgrade/migration callouts surfaced first. The ordering is deliberate: a developer gathers and de-noises the change inventory before any prose is written, a copywriter turns that inventory into benefit-led notes in a consistent voice, and a reviewer verifies every shipped change is represented and nothing internal-only or unreleased leaked in. Use this for GitHub Releases, product announcement posts, or the user-facing notes that ship alongside a tagged version.\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. Gather the change inventory (developer) — collect and de-noise every shipped change since the last tag → gated on artifact `{{workPath}}/gather.md` (markdown-notes)\n2. Write the notes (copywriter) — turn the inventory into benefit-led, user-facing notes → gated on `release-notes.md` (markdown-doc)\n3. Review for completeness and tone (reviewer) — verify coverage, accuracy, and that nothing internal leaked → gated on artifact `{{workPath}}/review.md` (markdown-notes)\n\nThe gates never advance with an unmet criterion, and loop back to the\nowning phase to fix named gaps.\n",
+  "entryStepId": "gather",
+  "triggers": [
+    "write release notes",
+    "release notes for v2",
+    "what changed this release",
+    "draft release announcement",
+    "notes for the new version",
+    "cut a changelog",
+    "update CHANGELOG.md",
+    "roll unreleased into a version",
+    "stamp the changelog for v1.4",
+    "keep a changelog release",
+    "write a changelog",
+    "generate release notes",
+    "whats new writeup",
+    "changelog from commits",
+    "version notes"
+  ],
+  "steps": [
+    {
+      "id": "gather",
+      "name": "Gather the change inventory",
+      "description": "collect and de-noise every shipped change since the last tag",
+      "prompt": "1. Determine the version range: the last released tag (or last release date) up to HEAD/the release candidate. 2. Collect the raw inventory — merged PRs, commit subjects, and linked issue IDs (use git log, gh pr list --state merged, or the changelog source available to you). 3. Filter OUT noise that should never reach users: merge commits, dependency-bump-only commits, CI/build-config tweaks, lint/format-only commits, and anything not yet actually released. 4. Classify each remaining change into exactly one bucket: Features, Improvements, Bug Fixes, Breaking Changes. 5. Flag every Breaking Change and anything requiring a user action (config change, migration, new env var) — these get top billing later. 6. Good looks like: a deduplicated, categorized list where each item has a PR/issue reference and a one-line plain description, and every category is either populated or explicitly marked 'none'. Write an acceptance-criteria checklist (e.g. 'every shipped user-facing change appears exactly once', 'breaking changes are flagged with the required user action', 'no merge/dependency/CI-only noise', 'each item cites a PR or issue') plus the version range and the classified inventory to the notes file via write_task_note AND to {{workPath}}/gather.md.\n\nThe deliverable `{{workPath}}/gather.md` lands in the project's artifacts drawer — write it with `write_artifact` and read it back with `read_artifact`; the shipped workspace stays untouched.",
+      "suggestedRole": "developer",
+      "advanceWhen": {
+        "file": "{{workPath}}/gather.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/gather.md",
+            "bytes": 120,
+            "artifact": true
+          },
+          {
+            "kind": "sniff",
+            "file": "{{workPath}}/gather.md",
+            "sniff": "nonempty",
+            "artifact": true
+          }
+        ],
+        "onReject": "gather",
+        "maxAttempts": 3
+      },
+      "next": "write"
+    },
+    {
+      "id": "write",
+      "name": "Write the notes",
+      "description": "turn the inventory into benefit-led, user-facing notes",
+      "prompt": "1. Open {{workPath}}/gather.md and write the release notes for HUMANS, not for the commit log — lead each item with the user-visible benefit, not the implementation. 2. Order the document: a one-paragraph highlights summary, then Breaking Changes / Upgrade Notes (with the exact action the user must take), then Features, Improvements, Bug Fixes. 3. Keep each line tight and parallel in voice; use the imperative or present tense consistently ('Adds…', 'Fixes…'). 4. Preserve the PR/issue references as links or trailing (#123) so readers can dig in. 5. Open with the version number and release date as an H1/H2. 6. Good looks like: a reader who has never seen the codebase understands what changed and what they must do, breaking changes are impossible to miss, and every inventory item from gather.md is represented. Write the notes to release-notes.md.\n\nBefore working, open the artifact input `{{workPath}}/gather.md` with `read_artifact`; do not look for it in the workspace.",
+      "suggestedRole": "copywriter",
+      "consumes": [
+        {
+          "file": "{{workPath}}/gather.md",
+          "artifact": true
+        }
+      ],
+      "advanceWhen": {
+        "file": "release-notes.md",
+        "minBytes": 1,
+        "sniff": "nonempty"
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "release-notes.md",
+            "bytes": 500
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkContains",
+            "scope": "standard",
+            "inputs": {
+              "file": "release-notes.md",
+              "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+              "flags": "i"
+            }
+          }
+        ],
+        "onReject": "write",
+        "maxAttempts": 3
+      },
+      "next": "review"
+    },
+    {
+      "id": "review",
+      "name": "Review for completeness and tone",
+      "description": "verify coverage, accuracy, and that nothing internal leaked",
+      "prompt": "1. Cross-check release-notes.md against {{workPath}}/gather.md: every classified item must appear, and nothing must appear that was filtered as noise or is unreleased. 2. Verify Breaking Changes are at the top with a concrete user action, not buried. 3. Check the voice is consistent and benefit-led, version + date are present, and references resolve. 4. Confirm no internal-only details (private ticket chatter, internal service names, secrets, profanity) leaked into a public document. 5. Write a PASS/FAIL against each acceptance criterion from gather.md, listing any specific gap. 6. If anything fails, the fix happens back in the write phase. Write your findings to {{workPath}}/review.md.\n\nBefore working, open the artifact input `{{workPath}}/gather.md` with `read_artifact`; do not look for it in the workspace.\n\nThe deliverable `{{workPath}}/review.md` lands in the project's artifacts drawer — write it with `write_artifact` and read it back with `read_artifact`; the shipped workspace stays untouched.",
+      "suggestedRole": "reviewer",
+      "consumes": [
+        {
+          "file": "{{workPath}}/gather.md",
+          "artifact": true
+        },
+        {
+          "file": "release-notes.md"
+        }
+      ],
+      "advanceWhen": {
+        "file": "{{workPath}}/review.md",
+        "minBytes": 1,
+        "sniff": "nonempty",
+        "artifact": true,
+        "requireChange": true
+      },
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "release-notes.md",
+            "bytes": 500
+          },
+          {
+            "kind": "citationsResolve",
+            "file": "{{workPath}}/review.md",
+            "minCitations": 2,
+            "artifact": true
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkContains",
+            "scope": "standard",
+            "inputs": {
+              "file": "release-notes.md",
+              "pattern": "(?:^|\\n)#{1,3}\\s+\\S",
+              "flags": "i"
+            }
+          }
+        ],
+        "onReject": "review",
+        "maxAttempts": 4
+      },
+      "next": "evaluate"
+    },
+    {
+      "id": "evaluate",
+      "name": "Evaluate",
+      "description": "Grade the deliverable against every acceptance criterion. All pass → finish; any fail → loop back and fix the gap.",
+      "prompt": "Review `{{workPath}}/review.md` against the original scope: every claim cites a real path, the severities and priorities are defensible, and nothing important in scope was skipped. Open the underlying files for at least the top findings; do not grade from the report's own summary. If a finding cannot be verified, say so explicitly.\n\nWrite your review to the artifact `{{workPath}}/review.md` with one `write_artifact` call: what you checked, what held, what did not (as a findings table with columns `| Severity | File | Line | Problem | Fix |` when anything needs fixing), ending with exactly `Verdict: PASS` or `Verdict: REVISE`. The gate enforces the verdict: a well-formed REVISE is routed back to the `review` step automatically with your findings. On PASS, `advance_task_step` to `finish`.",
+      "suggestedRole": "reviewer",
+      "consumes": [
+        {
+          "file": "release-notes.md"
+        }
+      ],
+      "next": "finish",
+      "gate": {
+        "at": "completion",
+        "checks": [
+          {
+            "kind": "minBytes",
+            "file": "{{workPath}}/review.md",
+            "bytes": 300,
+            "artifact": true
+          },
+          {
+            "kind": "contains",
+            "file": "{{workPath}}/review.md",
+            "pattern": "Verdict:\\s*(?:PASS|REVISE)",
+            "flags": "i",
+            "label": "explicit PASS or REVISE verdict",
+            "artifact": true
+          }
+        ],
+        "scripts": [
+          {
+            "name": "checkFixReview",
+            "scope": "standard",
+            "inputs": {
+              "taskRef": "{{task.ref}}",
+              "reviewPath": "{{workPath}}/review.md",
+              "fixStepId": "review"
+            }
+          }
+        ],
+        "onReject": "evaluate",
+        "maxAttempts": 4
+      }
+    },
+    {
+      "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": "1.0.3",
+  "releasedAt": "2026-08-24T00:00:00.000Z",
+  "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": "1.26236"
+}

package/data/craftbook-templates/re/release-notes/versions/1.0.3/test.json

@@ -0,0 +1,114 @@
+{
+  "schemaVersion": 1,
+  "title": "Public release notes from noisy change inventory",
+  "objective": "Measure whether the release-notes craftbook gathers, de-noises, classifies, and writes benefit-led release notes with breaking changes up front.",
+  "tags": [
+    "corpus"
+  ],
+  "prompt": "In the `Release Notes Eval` project, read `source/change-inventory.md` and write `release-notes.md` for HarborCRM 3.8.0 dated 2026-06-28. Include Highlights, Breaking Changes / Upgrade Notes, Features, Improvements, and Bug Fixes. Preserve PR references, exclude internal-only noise, and make the breaking config action impossible to miss.",
+  "setup": {
+    "projectName": "Release Notes Eval",
+    "about": "Self-contained eval project for release-notes. The raw change inventory is seeded in workspace/source/change-inventory.md.",
+    "missionObjectives": "Write public release notes from a noisy change inventory, preserving user-facing changes and references while excluding internal-only noise.",
+    "files": [
+      {
+        "path": "source/change-inventory.md",
+        "content": "# HarborCRM 3.8.0 change inventory\n\n- PR #812 Feature: adds saved invoice filters for finance teams.\n- PR #817 Improvement: webhook retry queue now shows next retry time in the admin panel.\n- PR #823 Breaking: integrations must rename env var `HC_WEBHOOK_SECRET` to `PAYMENTS_WEBHOOK_SECRET` before deploy.\n- PR #829 Bug fix: fixes duplicate customer merge suggestions after CSV import.\n- PR #831 Internal: renamed reducer files in the billing package.\n- PR #832 CI-only: upgraded lint runner and prettier config.\n- PR #835 Dependency bump only: patch-level dev dependency update.\n"
+      }
+    ],
+    "worker": {
+      "name": "Rina",
+      "role": "Copywriter"
+    },
+    "craftbookParams": {
+      "workPath": "tasks/eval"
+    }
+  },
+  "mocks": [],
+  "success": {
+    "summary": "workspace/release-notes.md has version/date, public sections, PR references, all user-facing changes, a prominent breaking action, and excludes internal/CI/dependency noise.",
+    "deliverables": [
+      {
+        "path": "release-notes.md",
+        "kind": "markdown-doc",
+        "minBytes": 1500,
+        "checks": [
+          {
+            "kind": "contains",
+            "file": "release-notes.md",
+            "pattern": "HarborCRM|3\\.8\\.0|2026-06-28"
+          },
+          {
+            "kind": "contains",
+            "file": "release-notes.md",
+            "pattern": "Highlights|Breaking Changes|Upgrade Notes|Features|Improvements|Bug Fixes",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "release-notes.md",
+            "pattern": "#812|#817|#823|#829",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "release-notes.md",
+            "pattern": "saved invoice filters|retry queue|duplicate customer merge",
+            "flags": "i"
+          },
+          {
+            "kind": "contains",
+            "file": "release-notes.md",
+            "pattern": "HC_WEBHOOK_SECRET|PAYMENTS_WEBHOOK_SECRET|before deploy|breaking",
+            "flags": "i"
+          },
+          {
+            "kind": "notContains",
+            "file": "release-notes.md",
+            "pattern": "reducer|lint runner|prettier|dependency bump|#831|#832|#835",
+            "flags": "i"
+          }
+        ]
+      }
+    ]
+  },
+  "rubric": {
+    "artifact": {
+      "path": "release-notes.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": [
+    "de-noising",
+    "benefit-led notes",
+    "breaking action visibility"
+  ],
+  "extensions": {
+    "legacySimulators": [
+      {
+        "id": "release-notes-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/re/release-pipeline-ci/manifest.json

@@ -3,6 +3,7 @@
   "kind": "craftbook-template",
   "id": "release-pipeline-ci",
   "role": "maintenance-review",
+  "category": "code-release",
   "name": "Release Pipeline (CI/CD)",
   "description": "Author a CI/CD release workflow as a YAML pipeline (GitHub Actions style) that takes a tagged commit and turns it into a published release — running tests and build as gates, then on a version tag building the artifact, generating release notes, publishing the package/image, and creating the GitHub Release, with the dangerous publish steps gated behind the green test/build stages and the tag trigger. A planner scopes the trigger, stages, secrets, and the publish target, a developer writes the valid pipeline YAML wiring the stages in the safe order, and a reviewer validates the syntax and that publishing cannot run before tests pass. Use this to set up automated releases, a publish-on-tag workflow, or a deploy pipeline from scratch.",
   "tags": [
@@ -19,6 +20,10 @@
   },
   "logo": "logo.webp",
   "license": "MIT",
-  "yankedVersions": [],
+  "yankedVersions": [
+    "1.0.0",
+    "1.0.1",
+    "1.0.2"
+  ],
   "workflow": "build-loop"
 }