@friedbotstudio/create-baseline 0.6.0 → 0.7.0

package/README.md

@@ -41,7 +41,7 @@ A discipline layer for Claude Code. Hooks at every tool boundary, a workflow tha
 
 ## What this is
 
-The Claude Code Baseline is a repository overlay shipped via `npx @friedbotstudio/create-baseline ./target`. It installs **22 hooks** at Claude's tool boundaries, **38 skills** organised into ten categories, **1 subagent** for parallel work in isolated worktrees, an **11-phase workflow** from intake to commit, and **3 user-typed consent gates** that Claude cannot forge.
+The Claude Code Baseline is a repository overlay shipped via `npx @friedbotstudio/create-baseline ./target`. It installs **22 hooks** at Claude's tool boundaries, **38 skills** organised into ten categories, **1 subagent** for parallel work in isolated worktrees, **4 canonical workflow tracks** declared in `.claude/workflows.jsonl` (where `intake-full` runs 11 nodes from intake to commit), and **3 user-typed consent gates** that Claude cannot forge.
 
 Every soft engineering rule a team usually repeats every session — *don't push, don't `--amend`, don't self-approve specs, don't skip phases, don't mock internal modules* — becomes a structural guarantee because the hooks run **outside Claude's tool boundary**. Claude cannot disable a hook with a flag, cannot write a consent marker, cannot reorder the phases without an explicit exception that triage records on disk.
 
@@ -62,9 +62,9 @@ A team that installs the baseline stops typing *"don't push, don't `--amend`, do
 | What | Count | Where it lives |
 |---|---:|---|
 | **Hooks** at PreToolUse / PostToolUse / SessionStart / Stop / PreCompact / UserPromptSubmit | 22 | `.claude/hooks/` |
-| **Skills** across artifact drafting, workflow phases, phase workers, spec helpers, orchestration, memory, audit, alternate tracks, and shared globals | 36 | `.claude/skills/` |
+| **Skills** across artifact drafting, workflow phases, phase workers, spec helpers, orchestration, memory, audit, alternate tracks, shared globals, and maintenance | 38 | `.claude/skills/` |
 | **Subagent** — `swarm-worker`, executes pre-decided recipes inside isolated git worktrees | 1 | `.claude/agents/` |
-| **Workflow phases** — intake → scout → research → spec → tdd → simplify → security → integrate → document → archive → memory-flush → commit | 11 | enforced by `track_guard` |
+| **Workflow tracks** declared in `.claude/workflows.jsonl`. Canonical set: `intake-full` (11 nodes), `spec-entry`, `tdd-quickfix`, `chore`. Two sub-tracks (`swarm-implementation`, `tdd-worker-chain`) are referenced by selector nodes inside the canonical set. | 4 selectable + 2 sub | `.claude/workflows.jsonl`, enforced by `track_guard` |
 | **Consent gates** — `/approve-spec`, `/approve-swarm`, `/grant-commit`. User-typed; structurally un-invokable by Claude | 3 | `consent_gate_grant` UserPromptSubmit hook |
 | **MCP servers** declared in `.mcp.json` — `context7` (third-party API docs), `plantuml` (diagram render), `playwright` (cross-engine smoke) | 3 | `.mcp.json` |
 
@@ -92,15 +92,19 @@ npx @friedbotstudio/create-baseline ./your-project
 npx @friedbotstudio/create-baseline ./your-project
 
 # Force-overwrite an existing install (interactive — type 'overwrite')
-npx @friedbotstudio/create-baseline ./your-project --overwrite
+npx @friedbotstudio/create-baseline ./your-project --force
 
 # Upgrade an existing install against a newer baseline version.
-# In a TTY, each customised file becomes a keep-mine / take-theirs / abort
-# prompt. In CI / piped stdout, reproduces the prior --merge behaviour:
+# In a TTY, each tier-1 customised file becomes a keep-mine / take-theirs / abort
+# prompt; tier-2 files auto-merge via `git merge-file --diff3`; tier-3 files
+# stage for the /upgrade-project Claude Code skill to reconcile. In CI / piped
+# stdout, every per-file action is reported with a user-facing label:
 #   - adds new baseline files
 #   - refreshes baseline files the user has not touched
-#   - preserves user-customised files (exit 3 if any)
-#   - removes baseline files the upstream removed (only if untouched locally)
+#   - keeps customised files (exit 3 if any preserved)
+#   - removes baseline files removed upstream that the user had not touched
+#   - exit 4 if a mechanical merge produced conflict markers
+#   - exit 5 if any tier-3 file was staged for /upgrade-project
 npx @friedbotstudio/create-baseline upgrade ./your-project
 
 # Preview without writing anything
@@ -158,7 +162,7 @@ The three workflow-phase consent gates pause the workflow until you type the cor
 
 - **`/approve-spec <slug>`** — after the spec phase, before any code is written
 - **`/approve-swarm <slug>`** — after `/swarm-plan`, before parallel dispatch
-- **`/grant-commit`** — after `/archive`, before the commit lands
+- **`/grant-commit`** — after `/archive` and `/memory-flush`, before the commit lands
 
 A fourth consent gate sits outside the phase pipeline:
 
@@ -172,7 +176,7 @@ The 22 hooks declared in `.claude/settings.json` fire at every Claude tool bound
 
 The architectural rule is simple: **decisions live in main context; subagents only execute pre-decided recipes**. The baseline ships exactly one subagent — `swarm-worker` — and its only sanctioned use is parallel dispatch of fully-specified recipes inside isolated git worktrees during `/swarm-dispatch`. Every other capability that might have been a subagent (code authoring, scenario design, scouting, security review, prose writing, UI design) lives instead as a **skill** that runs in main context with full conversation visibility.
 
-The 11-phase workflow is enforced at the write boundary by `track_guard`. Phase ordering is binding; the only mechanism to bypass a phase is the `exceptions` array in `.claude/state/workflow.json`, written by `/triage` at workflow creation time. The chore track is a stripped-down ordering of the same gates, with the test-first phases removed because there is nothing to test-first.
+Tracks declared in `.claude/workflows.jsonl` are enforced at the write boundary by `track_guard`. Node ordering inside each track is binding; the only mechanism to bypass a node is the `exceptions` array in `.claude/state/workflow.json`, written by `/triage` at workflow creation time. The `chore` track is a stripped-down ordering of the same gates, with the test-first nodes removed because nothing needs testing first. Projects declare their own tracks (or add nodes to the canonical ones) by editing their `.claude/workflows.jsonl`. Article IV's invariants (I1..I11) bind every track regardless of who wrote it; a track that omits `/grant-commit` before a `commit` node, or whose dependency graph contains a cycle, is rejected at triage time with a named error.
 
 The constitution at `CLAUDE.md` is the source of truth for in-session behaviour; `docs/init/seed.md` is the source of truth for the baseline's shape. When the constitution and the implementation conflict, the constitution governs and the implementation gets corrected. When `seed.md` and the constitution conflict, `seed.md` governs and you stop and surface the drift before acting.
 

package/bin/cli.js

@@ -8,8 +8,8 @@ import { existsSync } from 'node:fs';
 import * as io from '../src/cli/io.js';
 import { scanSentinels } from '../src/cli/conflict.js';
 import { freshInstall, forceInstall, COPY_EXCLUDE } from '../src/cli/install.js';
-import { threeWayMerge } from '../src/cli/merge.js';
-import { loadManifest, buildManifestFromDir } from '../src/cli/manifest.js';
+import { threeWayMerge, ACTION_LABELS, ACTION_LABEL_WIDTH } from '../src/cli/merge.js';
+import { loadManifest, buildManifestFromDir, MANIFEST_VERSION } from '../src/cli/manifest.js';
 import { fetchPlantumlIfMissing, FETCH_OUTCOMES } from '../src/cli/plantuml.js';
 import { runDoctor, formatReport } from '../src/cli/doctor.js';
 
@@ -66,11 +66,13 @@ Misc:
   --version        Print version.
 
 Exit codes:
-  0  success / clean doctor
+  0  success (or clean doctor)
   1  user abort, conflict-without-force, doctor reports missing files, or upgrade aborted
-  2  argv error, non-TTY where TTY required, doctor finds no manifest, or --merge passed
-  3  upgrade had skipped customizations (or stale-customized prunes)
-  4  --require-plantuml fetch failure
+  2  bad command line, prompted input needed but no terminal, doctor finds no manifest,
+     or deprecated --merge passed
+  3  upgrade kept your customized files (or preserved files removed upstream)
+  4  --require-plantuml jar fetch failed, OR mechanical merge produced conflicts to resolve
+  5  semantic merge staged for /upgrade-project to reconcile in Claude Code
 `;
 
 const OPTIONS = {
@@ -203,8 +205,9 @@ async function runPlainInstall(target, values, templateDir) {
   if (!dryRun) {
     const plantumlExit = await fetchPlantumlPlain(target, values);
     if (plantumlExit !== 0) return plantumlExit;
-    io.log(`Installed manifest version 1 to ${target}.`);
-    io.log(`Pin via "@friedbotstudio/create-baseline@<exact-version>" in your bootstrap docs.`);
+    const version = await readPackageVersion();
+    io.log(`Baseline installed at ${target} (manifest v${MANIFEST_VERSION}).`);
+    io.log(`For reproducible installs, pin the exact version: @friedbotstudio/create-baseline@${version}`);
   }
   return 0;
 }
@@ -232,11 +235,11 @@ async function dispatchUpgrade(target, values, templateDir) {
     await usageError(`No baseline manifest at ${manifestPath}. Run a fresh install first.`);
     return 2;
   }
-  const { findPendingStage } = await import('../src/cli/upgrade-tiers.js');
+  const { findPendingStage, formatStageTimestamp } = await import('../src/cli/upgrade-tiers.js');
   const pending = await findPendingStage(target);
   if (pending) {
     const fileLines = pending.files.map((f) => `  - ${f}`).join('\n');
-    io.log(`Pending semantic-merge stage at ${pending.stage_ts}.\n${pending.files.length} file(s) awaiting reconciliation:\n${fileLines}\nOpen Claude Code and run /upgrade-project to reconcile.`);
+    io.log(`A previous upgrade staged ${pending.files.length} file(s) for Claude Code review (staged ${formatStageTimestamp(pending.stage_ts)}):\n${fileLines}\nOpen Claude Code and run /upgrade-project to reconcile.`);
     return 5;
   }
   if (process.stdout.isTTY) {
@@ -260,7 +263,8 @@ async function runPlainUpgrade(target, values, templateDir, manifestPath) {
   }
   const report = await threeWayMerge(templateDir, target, oldManifest, newManifest);
   for (const action of report.actions) {
-    io.log(`${action.kind.padEnd(28)} ${action.path}`);
+    const label = ACTION_LABELS[action.kind] ?? action.kind;
+    io.log(`${label.padEnd(ACTION_LABEL_WIDTH)}  ${action.path}`);
   }
   return report.exitCode;
 }
@@ -353,7 +357,7 @@ async function main(argv) {
   }
 
   if (values['no-plantuml'] && values['require-plantuml']) {
-    await usageError('--no-plantuml and --require-plantuml are mutually exclusive');
+    await usageError('--no-plantuml and --require-plantuml conflict; pick one or omit both.');
     return 2;
   }
   if (positionals.length === 0) {

package/obj/template/.claude/commands/init-project-doctor.md

@@ -0,0 +1,74 @@
+---
+description: Detect and repair baseline drift — missing or invalid `.claude/workflows.jsonl`, schema violations, four-way Article IV / §18 mirror drift, and (advisory) shipped-tooling files placed outside `.claude/` against the convention codified at seed.md §3. Interactive: presents each detected violation via AskUserQuestion and applies the named fix on confirmation.
+argument-hint: ""
+allowed-tools: Read, Write, Edit, Bash, AskUserQuestion, Glob, Grep
+disable-model-invocation: true
+---
+
+# `/init-project doctor` — baseline drift detector + repairer
+
+User-only command. Run after a `create-baseline upgrade` cycle, after manually editing `workflows.jsonl`, or any time the baseline feels out of sync. Different from `create-baseline doctor` (the npm CLI manifest-drift checker) — this is a project-internal interactive repair tool.
+
+## Step 1 — Pre-flight
+
+- Print: "`/init-project doctor` — checking baseline integrity. Each detected violation is presented separately; you confirm each fix via AskUserQuestion."
+- Verify the project is configured: `.claude/project.json → configured == true`. If false, halt with: "Run `/init-project` first; the project is in agnostic mode."
+- Verify `python3` and `node` available; halt with one-line missing-dep message if not.
+
+## Step 2 — Check `.claude/workflows.jsonl` presence
+
+- If `.claude/workflows.jsonl` is missing on disk:
+  - AskUserQuestion: "`.claude/workflows.jsonl` missing. Restore from pristine template (`src/.claude/workflows.template.jsonl`)?"
+  - Options: `Restore` | `Skip` | `Show diff`
+  - On `Restore`: `cp src/.claude/workflows.template.jsonl .claude/workflows.jsonl`. Log to `.claude/state/init/doctor-<timestamp>.log`.
+  - On `Skip`: note in report; proceed.
+
+## Step 3 — Check `.claude/schemas/` presence
+
+- If `.claude/schemas/workflow-track.v1.json` is missing:
+  - AskUserQuestion: "`.claude/schemas/workflow-track.v1.json` missing. Restore from `src/.claude/schemas/`?"
+  - Apply on `Restore` (recursive cp).
+
+## Step 4 — Validate workflows.jsonl against §18 schema + invariants
+
+- Run `node .claude/skills/triage/seed-tasklist.mjs --validate-only`. The helper exits 0 on success or non-zero with a named-error report on failure.
+- On validation failure, parse the helper's stderr for each error. For each:
+  - AskUserQuestion: "Violation: `<kind>` in track `<id>`: `<message>`. Options: `Show context`, `Skip` (mark as known), `Edit manually` (open file at line)."
+  - The doctor does NOT auto-fix schema/invariant violations — manual user judgment is required. Surface the violation context (track + node ids) and pause.
+
+## Step 5 — Four-way mirror check (Article IV / §18)
+
+Extract the §18 sections from `docs/init/seed.md` and `src/seed.template.md`. Extract Article IV sections from `CLAUDE.md` and `src/CLAUDE.template.md`. Byte-compare each pair.
+
+- On §18 mirror drift:
+  - AskUserQuestion: "`docs/init/seed.md §18` differs from `src/seed.template.md §18`. Options: `Re-mirror docs→src`, `Re-mirror src→docs`, `Show diff`, `Skip`."
+  - Apply the chosen overwrite.
+
+- Same pattern for Article IV mirror.
+
+## Step 6 — `.claude/` tooling convention check (advisory)
+
+Per seed.md §3 + `conventions.md → user-shipped-tooling-lives-in-claude-directory`, user-shipped baseline tooling lives under `.claude/`. The only project-root exceptions are `CLAUDE.md` and `.mcp.json`.
+
+- Scan for shipped tooling at the project root that should live under `.claude/`. Heuristic: files matching `*.skill.md`, `*.workflow.json`, `*.hook.sh`, `*.command.md` at the project root.
+- For each match:
+  - AskUserQuestion: "Convention violation: `<path>` at project root. Suggested target: `.claude/<subdir>/<filename>`. Move?"
+  - The doctor offers the move; the user confirms.
+
+## Step 7 — Report + log
+
+- Print a summary table:
+  - Checks run: N
+  - Fixes applied: M
+  - Skipped (user declined or manual fix required): K
+  - Remaining manual: L
+- Write the full session record to `.claude/state/init/doctor-<UTC-timestamp>.log` (one line per check + action).
+- If any check returned a manual-fix path (e.g., schema violation in workflows.jsonl), exit with code 1 so caller workflows treat the result as "drift remains". Otherwise exit 0.
+
+## Constraints
+
+- **Read-only by default; writes only on user confirmation.** Every Edit/Write happens after an AskUserQuestion `Apply` response. No silent fixes.
+- **No commits.** The doctor's fixes land on the working tree. The user commits via the normal `/grant-commit` + `/commit` flow.
+- **No new dependencies.** Validation reuses `src/cli/workflows-validator.js` via the triage helper.
+- **Schema/invariant violations require manual fixes.** Auto-fixing structurally violates the user's intent (the violation might be intentional during development). Doctor surfaces; user fixes.
+- **Mirror drift CAN be auto-fixed** (one-direction overwrite on confirmation). The reverse direction (which file is canonical?) is the user's call.

package/obj/template/.claude/hooks/lib/resume_writer.py

@@ -191,7 +191,20 @@ def compose_snapshot(transcript: Path, project_dir: Path, trigger: str) -> str:
     workflow = _read_workflow(project_dir)
 
     slug = workflow.get("slug") or "(none)"
-    entry_phase = workflow.get("entry_phase") or "(unknown)"
+    # Post-§18: workflow.json carries `track_id`; legacy pre-§18 files carry
+    # `entry_phase`. Accept both. Map reverses
+    # `src/cli/workflow-migrator.js → ENTRY_PHASE_TO_TRACK_ID`.
+    _TRACK_ID_TO_ENTRY_PHASE = {
+        "intake-full": "intake",
+        "spec-entry": "spec",
+        "tdd-quickfix": "tdd",
+        "chore": "chore",
+    }
+    entry_phase = (
+        workflow.get("entry_phase")
+        or _TRACK_ID_TO_ENTRY_PHASE.get(workflow.get("track_id"))
+        or "(unknown)"
+    )
     completed = workflow.get("completed") or []
     exceptions = workflow.get("exceptions") or []
     phases = workflow.get("phases") or []

package/obj/template/.claude/hooks/track_guard.sh

@@ -54,7 +54,17 @@ except Exception:
 phases = pj.get("workflow", {}).get("phases") or []
 artifacts = pj.get("workflow", {}).get("artifacts") or {}
 exceptions = set(ws.get("exceptions") or [])
-entry = ws.get("entry_phase")
+# Post-§18: workflow.json carries `track_id`; legacy pre-§18 files carry
+# `entry_phase`. Accept both. The canonical map below mirrors
+# `src/cli/workflow-migrator.js → ENTRY_PHASE_TO_TRACK_ID` in reverse so
+# track_guard's phase-ordering enforcement keeps working on both shapes.
+_TRACK_ID_TO_ENTRY_PHASE = {
+    "intake-full": "intake",
+    "spec-entry": "spec",
+    "tdd-quickfix": "tdd",
+    "chore": "chore",
+}
+entry = ws.get("entry_phase") or _TRACK_ID_TO_ENTRY_PHASE.get(ws.get("track_id"))
 
 # Find which phase this file belongs to.
 file_phase = None

package/obj/template/.claude/manifest.json

@@ -1,6 +1,6 @@
 {
   "manifest_version": 3,
-  "generated_at": "2026-05-20T12:52:27.375Z",
+  "generated_at": "2026-05-21T16:49:56.519Z",
   "files": {
     ".claude/agents/swarm-worker.md": {
       "sha256": "1735a220f268c9765cb22e0567b728803f2edd7776cbde51dd017a9f062ae41f",
@@ -30,6 +30,10 @@
       "sha256": "770e8bc3bf1097ecdf92b86dcf6e0cf7ef3a3251b73bf35cc0385f2e2d552319",
       "tier": "MECHANICAL"
     },
+    ".claude/commands/init-project-doctor.md": {
+      "sha256": "c71bcc88aaf77095dd87ffc88a1f79ad731607623240cfb9125bf8b074425897",
+      "tier": "MECHANICAL"
+    },
     ".claude/commands/init-project.md": {
       "sha256": "222f02f65ea639c904c43114642f11422a56d8dab8c1c2786b5876e17d3b0155",
       "tier": "MECHANICAL"
@@ -67,7 +71,7 @@
       "tier": "MECHANICAL"
     },
     ".claude/hooks/lib/resume_writer.py": {
-      "sha256": "9592fb72de966cb797e51322fbf0160a6784de34934bd6587768689a34bd03b5",
+      "sha256": "2811861400c3afbfcb57df6c0889676abda0716e5acf35a2b0ba9410d44d6d45",
       "tier": "MECHANICAL"
     },
     ".claude/hooks/lint_runner.sh": {
@@ -151,7 +155,7 @@
       "tier": "MECHANICAL"
     },
     ".claude/hooks/track_guard.sh": {
-      "sha256": "cd79823bf4c24a35133ee6fc072e8c39bdfbc49042c698dcbeb2c36986f15c5f",
+      "sha256": "403f7cb1f29cbb7fb25ce96c5ea93a31f597fbc11ba7cfa2e313804110f44e75",
       "tier": "MECHANICAL"
     },
     ".claude/hooks/verify_pass_guard.sh": {
@@ -202,6 +206,10 @@
       "sha256": "a81f9c1341a15831fe8dd558e376dfe38d5fd938aac6291a4c59529637a8d2c7",
       "tier": "NEVER_TOUCH"
     },
+    ".claude/schemas/workflow-track.v1.json": {
+      "sha256": "d012a5eaea66b6172f05a1f9ed22ee14436e900372ed2653124e4a9ff1ef26e1",
+      "tier": "NEVER_TOUCH"
+    },
     ".claude/settings.json": {
       "sha256": "feb7d2c004b6c12cef6410d54db8ad0abc818b2aa16b1942cfdb72808c21b30d",
       "tier": "BINARY_PROMPT"
@@ -219,7 +227,7 @@
       "tier": "MECHANICAL"
     },
     ".claude/skills/audit-baseline/audit.sh": {
-      "sha256": "040335b18c73cb72b33865ccad4841cc9e9514174d2e09820ea3d822a45ad517",
+      "sha256": "6d799ea8b567ecf91757fa142f72795b1da0b4d8443e54b4d233dad5b98b5dab",
       "tier": "MECHANICAL"
     },
     ".claude/skills/audit-baseline/tests/fixtures/_pending_opener_only.md": {
@@ -307,7 +315,7 @@
       "tier": "MECHANICAL"
     },
     ".claude/skills/chore/SKILL.md": {
-      "sha256": "3479c4ef4c7706928b54ed89678c3526c790e2aeb629674bd60e5a9e4b90d80d",
+      "sha256": "88b2b8b29f0ab6ab4ad41fa823d3bd5a4d3f863b81e607ac69abbbd298eb6766",
       "tier": "MECHANICAL"
     },
     ".claude/skills/claude-automation-recommender/LICENSE": {
@@ -414,72 +422,8 @@
       "sha256": "456160c1a3398cfece91b6696d0c7f1fcb1196ea5e101acb9a0e3050f2462a5d",
       "tier": "MECHANICAL"
     },
-    ".claude/skills/google-analytics/SKILL.md": {
-      "sha256": "f5d8c52c103e4e1bdbb6bfdda962aad198064ced701c20df753d77ba4603547c",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/audiences.md": {
-      "sha256": "cdae62b0f40e82a60f96a6bd0e77798c6118beaebe812961e8c9d4c3242b227f",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/bigquery.md": {
-      "sha256": "fadc3efdd200a0bc43df4b698d46bc93e82111d8553b363d1e49ceb15bcdffce",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/custom-dimensions.md": {
-      "sha256": "ac9e88c69fd8eb2a0bf1ae8f3daac42c6d92caa338ce2fa77a50bf2fe947dbc4",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/custom-events.md": {
-      "sha256": "442ce8c1e7d89cb2206a85b02b287748e3e226d67cab571bc6f2a3ea520bc434",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/data-management.md": {
-      "sha256": "1f9e80e0855937dc082d63535188fc1e7547273e626d0a3ef03c0487b42045bb",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/debugview.md": {
-      "sha256": "f2b050db37b7927de99b0196deb7cb94d8ecb1a6668c9e08bec4f57816646e90",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/events-fundamentals.md": {
-      "sha256": "acdeb349be2f468ba634d686691aeda0ce932db93b4872b9cd35685dcd3f31cf",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/gtag.md": {
-      "sha256": "bef0556071bc645656d0634c26c3bad20d39517cb14689bb11e9d9bedf4f6ab8",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/gtm-integration.md": {
-      "sha256": "3c95c7b64fbbd688907cd558c3c4d265b99c7178284b97e376334d802d11c860",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/measurement-protocol.md": {
-      "sha256": "a5b644a5ca586173e399c229a47aad571991389232320a0f9e6baf1d4be619fb",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/privacy.md": {
-      "sha256": "09e2db4ce4d990f4f1ff51c0e3bca08a364cba711f83f48103f69f9db427fa18",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/recommended-events.md": {
-      "sha256": "0dccd2084060bfb3fd28476b9c0c9a6e2b8f51ce57ecf4fc32d7f82a8964b1b4",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/reporting.md": {
-      "sha256": "37cb7cfd1dd547e86a41664ffe13893ef138194a2288d1a2c43c9195e8ec609f",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/setup.md": {
-      "sha256": "26743b72341a5658e73c1f070b924de0f11b1681af06164cb13c19f2047a1087",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/google-analytics/references/user-tracking.md": {
-      "sha256": "5a945df2a575291c6b35d24830b497915c82a843a93ee133577f1deda93f6d75",
-      "tier": "MECHANICAL"
-    },
     ".claude/skills/harness/SKILL.md": {
-      "sha256": "af7cf374013d881d216462ccba353747e1727f670bd9b1c45b089ffde247d67a",
+      "sha256": "9ddcc243fc72c48c205dc688094d6305f7969969b4b0f899bab428e1abdb1ce3",
       "tier": "MECHANICAL"
     },
     ".claude/skills/humanizer/LICENSE": {
@@ -723,7 +667,7 @@
       "tier": "MECHANICAL"
     },
     ".claude/skills/intake/SKILL.md": {
-      "sha256": "7ad00f23ee28737dccef9c71ae477a54edad34a09f50b7774a34ae11e2ca016b",
+      "sha256": "7f73e8ee3f3a5bc6e7e29f814e4820a09f2c12644d54ae45314dc733f188aab2",
       "tier": "MECHANICAL"
     },
     ".claude/skills/intake/template.md": {
@@ -746,26 +690,6 @@
       "sha256": "5eec4486cfda45d2ce5987d6286bf61b96f816ebdd9041bc9ba0207cf00abd4b",
       "tier": "MECHANICAL"
     },
-    ".claude/skills/optimize-seo/SKILL.md": {
-      "sha256": "b389dc604346220d5a0a59c8f41c058c192c58a35bbff831fbc4c7f9e8c1b386",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/optimize-seo/scripts/pagespeed.mjs": {
-      "sha256": "8bef62078833139f9175c175339d1fc0eb635a5eeaa2a3100543928ef91869e4",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/pagespeed-insights/LICENSE.md": {
-      "sha256": "10ee2ded8c2ae373a65aacfa51f2bf540d5f20d418fc7a5595a5ccb06cde2738",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/pagespeed-insights/SKILL.md": {
-      "sha256": "ab53af083929fbb28af0ad2c9ee1fcc51dec63c642ebc237c475b3ea9bde7c59",
-      "tier": "MECHANICAL"
-    },
-    ".claude/skills/pagespeed-insights/reference.md": {
-      "sha256": "afa03220b772ea8d6393f7366bbd6d90200c0b56585ccb2bb714e622af07345f",
-      "tier": "MECHANICAL"
-    },
     ".claude/skills/prose/SKILL.md": {
       "sha256": "7cfc7d4fd34d7256367eedf17c5f5b5cdb67baec72033f905dcfb715366a1849",
       "tier": "MECHANICAL"
@@ -839,7 +763,7 @@
       "tier": "MECHANICAL"
     },
     ".claude/skills/swarm-plan/SKILL.md": {
-      "sha256": "6e530d5b67151ad1a4d310a966fbf4e742cc077a34c931350603a05e26fb6e88",
+      "sha256": "cf0f60a10377e5913e327190c67ea7bf6ce978356c81b0572a3bd0078ffbe81e",
       "tier": "MECHANICAL"
     },
     ".claude/skills/swarm-plan/validate.sh": {
@@ -847,7 +771,7 @@
       "tier": "MECHANICAL"
     },
     ".claude/skills/tdd/SKILL.md": {
-      "sha256": "284a2f3737f3f25242affbb8bc04c03a48e1c151cbaa06198820184f213a4a36",
+      "sha256": "717ef6297934be3a1c2efb5fd6beb434dbca9eeeb4356ae7ada85c80555fde01",
       "tier": "MECHANICAL"
     },
     ".claude/skills/tdd/drift_check.py": {
@@ -891,7 +815,11 @@
       "tier": "MECHANICAL"
     },
     ".claude/skills/triage/SKILL.md": {
-      "sha256": "499f82a26d77c60af827dd89a7fcb3eae0cd903e34c70740c297bc260aeed38e",
+      "sha256": "6a5790c692ee03975a50956b2b1e0da7447f00c15378a0f0e69332403906e9b7",
+      "tier": "MECHANICAL"
+    },
+    ".claude/skills/triage/seed-tasklist.mjs": {
+      "sha256": "0459f1862f5c4cf500d5cf2b66dbcfa3874843306c10f5f915b0670999b0753b",
       "tier": "MECHANICAL"
     },
     ".claude/skills/upgrade-project/SKILL.md": {
@@ -902,16 +830,20 @@
       "sha256": "fcddba67cf7f3623fc8f8ae142303b16b9db0c9fc1652bc920e16c56fe4c7864",
       "tier": "MECHANICAL"
     },
+    ".claude/workflows.jsonl": {
+      "sha256": "e6a9bd01388b34a15758c8956df5986c917508d65bacef49a444079112da4090",
+      "tier": "NEVER_TOUCH"
+    },
     ".mcp.json": {
       "sha256": "8ebb7966045486187bbdf9bac643e690c4fbc7a9a70a8345e3665ba72fa19b96",
       "tier": "SPECIAL_MERGE"
     },
     "CLAUDE.md": {
-      "sha256": "18572ddaa929806897d17e6c0a76121eec622dced5d375c05fdd2788214571fd",
+      "sha256": "03023db4f879d3ee4ce37c8824676bab2cc999d4dd4d011bfdbe997f52c5f915",
       "tier": "SEMANTIC"
     },
     "docs/init/seed.md": {
-      "sha256": "150393a4486a25bfa301081c7328ca9785f6115c8211a8b8de27c2d7912def68",
+      "sha256": "7c64057392b39289edf123e3395b67b8605c28b9ab35c102c9469f6974047c78",
       "tier": "SEMANTIC"
     }
   },
@@ -957,5 +889,5 @@
       "verify": "baseline"
     }
   },
-  "build_id": "gha-26163688924"
+  "build_id": "gha-26240106828"
 }

package/obj/template/.claude/schemas/workflow-track.v1.json

@@ -0,0 +1,64 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "./workflow-track.v1.json",
+  "title": "Workflow Track v1",
+  "description": "One Track record per line in .claude/workflows.jsonl. The Article IV invariants I1..I11 are enforced at runtime by src/cli/workflows-validator.js, NOT by JSON Schema alone — schema validation is structural; invariant validation is semantic. See docs/init/seed.md §17 for the full semantic contract.",
+  "type": "object",
+  "required": ["$schema", "track_id", "name", "description", "selectable", "selector_hints", "preconditions", "invariants", "nodes"],
+  "additionalProperties": false,
+  "properties": {
+    "$schema": { "type": "string" },
+    "track_id": { "type": "string", "minLength": 1 },
+    "name": { "type": "string", "minLength": 1 },
+    "description": { "type": "string" },
+    "selectable": { "type": "boolean" },
+    "selector_hints": { "type": "array", "items": { "type": "string" } },
+    "preconditions": { "type": "array", "items": { "$ref": "#/$defs/Predicate" } },
+    "invariants": { "type": "array", "items": { "enum": ["commits", "requires_spec", "requires_swarm", "chore", "git_only"] } },
+    "nodes": { "type": "array", "minItems": 1, "items": { "$ref": "#/$defs/Node" } }
+  },
+  "$defs": {
+    "Predicate": {
+      "type": "object",
+      "required": ["name"],
+      "additionalProperties": false,
+      "properties": {
+        "name": { "enum": ["requires_git", "requires_user_override", "requires_min_components", "requires_phase_completed", "requires_skill_present"] },
+        "argument": { "type": "string" }
+      }
+    },
+    "Node": {
+      "type": "object",
+      "required": ["id", "type", "depends_on", "blocks", "can_parallel", "needs_user"],
+      "additionalProperties": false,
+      "properties": {
+        "id": { "type": "string", "minLength": 1 },
+        "type": { "enum": ["task", "selector"] },
+        "skill": { "type": "string" },
+        "sub_track": { "type": "string" },
+        "alternates": { "type": "array", "items": { "$ref": "#/$defs/Alternate" } },
+        "input": { "type": "string" },
+        "invocation_prompt": { "type": "string" },
+        "output": { "type": "string" },
+        "output_formatter_prompt": { "type": "string" },
+        "depends_on": { "type": "array", "items": { "type": "string" } },
+        "blocks": { "type": "array", "items": { "type": "string" } },
+        "can_parallel": { "type": "boolean" },
+        "needs_user": { "type": "boolean" },
+        "activeForm": { "type": "string" },
+        "metadata": { "type": "object" }
+      }
+    },
+    "Alternate": {
+      "type": "object",
+      "required": ["preconditions"],
+      "additionalProperties": false,
+      "properties": {
+        "skill": { "type": "string" },
+        "sub_track": { "type": "string" },
+        "preconditions": { "type": "array", "items": { "$ref": "#/$defs/Predicate" } },
+        "description": { "type": "string" }
+      }
+    }
+  }
+}

package/obj/template/.claude/skills/audit-baseline/audit.sh

@@ -80,7 +80,7 @@ def read_skill_owner(slug):
     m = re.search(r'^owner:\s*(\S+)\s*$', fm.group(1), re.MULTILINE)
     return m.group(1) if m else None
 
-EXPECTED_COMMANDS = {"approve-spec", "approve-swarm", "grant-commit", "grant-push", "init-project"}
+EXPECTED_COMMANDS = {"approve-spec", "approve-swarm", "grant-commit", "grant-push", "init-project", "init-project-doctor"}
 
 EXPECTED_MEMORY_FILES = {
     # Canonical files (seven)
@@ -186,7 +186,7 @@ skills_claimed  = find_count(
     r"\b(\d+|twenty-(?:four|five|six|seven|eight|nine)|"
     r"thirty|thirty-(?:one|two|three|four|five|six|seven|eight|nine)|forty)\s+skills?\b")
 gates_claimed   = find_count(r"\b(\d+|three)\s+consent\s+gates?\b")
-cmds_claimed    = 5 if re.search(r"four\s+consent\s+gates?\s*\+\s*one\s+bootstrap", seed, re.IGNORECASE) else None
+cmds_claimed    = 6 if re.search(r"four\s+consent\s+gates?\s*\+\s*one\s+bootstrap\s*\+\s*one\s+doctor", seed, re.IGNORECASE) else (5 if re.search(r"four\s+consent\s+gates?\s*\+\s*one\s+bootstrap", seed, re.IGNORECASE) else None)
 
 def check_count(label, claimed, actual):
     if claimed is None:

package/obj/template/.claude/skills/chore/SKILL.md

@@ -35,7 +35,7 @@ The classification rule is: *if there is no failing test that should exist for t
 
 ## Prereq
 
-`.claude/state/workflow.json` exists with `entry_phase: "chore"` (written by `/triage`). If the prereq is not met, refuse and surface the mismatch.
+`.claude/state/workflow.json` exists with `track_id == "chore"` (post-§18; legacy `entry_phase == "chore"` accepted for pre-§18 in-flight workflows). If the prereq is not met, refuse and surface the mismatch.
 
 ## Phase shape
 
@@ -74,7 +74,7 @@ If a conditional phase is required, run it **before** `/grant-commit`. If you sk
 
 ## Steps
 
-1. Read `.claude/state/workflow.json`. Confirm `entry_phase == "chore"`. If not, stop and surface the mismatch — the user reached this skill without the correct triage classification.
+1. Read `.claude/state/workflow.json`. Confirm `track_id == "chore"` (post-§18) OR `entry_phase == "chore"` (legacy pre-§18). If neither, stop and surface the mismatch — the user reached this skill without the correct triage classification.
 2. Restate the intended edits inline: file paths, brief description per file, estimated total diff size. Confirm with the user if anything is ambiguous.
 3. Apply the edits via `Edit` / `MultiEdit` / `Write`. Honour the engineering rules from CLAUDE.md Article VI (no stubs, no commented-out code, no `TODO` / `FIXME` / `HACK` / `XXX`).
 4. **Run the binding test command and stamp the verdict (inlined verify).** Per `.claude/skills/verify/SKILL.md` (the contract doc): read `.claude/project.json → test.cmd`; run via Bash from project root (capture stdout, stderr, exit code; no retry); apply verdict rules (`PASS` iff exit 0 AND at least one test executed AND no failed/errored test; otherwise `FAIL`); atomically write `.claude/state/last_test_result` with the canonical four-line format. The `verify_pass_guard` hook reads line 1 as the binding verdict. If the verdict is `FAIL`, stop — the user investigates; chore does not loop. Write `.claude/state/harness_state` with `state: "yielded"` and `reason: "chore verify FAIL"` so the Stop hook stays silent.