create-harness-vibe-coding 0.8.9 → 0.8.12

package/templates/common/Harness/WF-AUTO-SPARK.md

@@ -24,6 +24,8 @@ WF-AUTO-SPARK is a `/wf` variant layered on WF-AUTO: it inherits the WF-KERNEL c
 
 Accepted spark candidates re-enter the standard W2-W5 gates per [WF-KERNEL.md](WF-KERNEL.md): implementer (one file_claim, ≤3 files, ≤50 lines), verifier, reviewer(s), reflector. Spark searchers are read-only. Any implementation must use the dispatch packet from `Harness/dispatch.md` with explicit write set, forbidden truth files, AC IDs, and verification commands. A candidate with no verifier evidence or no reflector PASS is not accepted, even if the idea is valuable.
 
+For task-state writes, CEO provides concise bullets and decisions; task-scribe formats task-state writes for roadmap, cycle PLAN, heartbeat, evidence pointers, and closeout. If task-scribe is unavailable, CEO may write the smallest durable checkpoint and must record that degradation in `PROGRESS.md`. This keeps the high-reasoning controller focused on decisions instead of spending tokens formatting process files.
+
 ## Startup: Roadmap Declaration
 
 Before first spark cycle, CEO MUST declare a roadmap. This is written to `Harness/tasks/auto/SPARK-ROADMAP.md`.
@@ -89,12 +91,12 @@ STARTUP: Declare roadmap (North Star + Milestones) → user confirms
 │   ↓                                                     │
 │ VERIFY + CROSS-REVIEW + REFLECTOR PASS                  │
 │   ↓                                                     │
-│ VALUE REFLECTION: CEO writes in PROGRESS.md             │
+│ VALUE REFLECTION: CEO gives bullets; task-scribe writes PROGRESS.md │
 │   "Driven by [source]. Matters because [reason].        │
 │    Without this [consequence]. User notices [evidence]." │
 │   ↓                                                     │
 │ MILESTONE CHECK: Is current milestone complete?         │
-│   Yes → update SPARK-ROADMAP.md, celebrate in PROGRESS  │
+│   Yes → task-scribe updates SPARK-ROADMAP.md + PROGRESS │
 │   No  → continue                                       │
 │   ↓                                                     │
 │ RE-ANCHOR GATE (every 10 cycles):                       │
@@ -122,6 +124,13 @@ STARTUP: Declare roadmap (North Star + Milestones) → user confirms
 | 7 | **Developer Experience** | WebSearch / Docs | New tooling, better CLI patterns, improved error messages |
 | 8 | **Performance Benchmarks** | WebSearch | Industry benchmarks, optimization techniques, profiling tools |
 
+### Spark Search Fallbacks
+
+- If a researcher/search subagent is filtered, unavailable, or blocked by tool policy, CEO may run direct WebSearch/Tavily for discovery only and records `searchFallback: ceo-direct`.
+- If a scoped skill name is unavailable, use the unscoped skill name plus a dispatch packet; if that still fails, continue with direct search and record the failure.
+- Discovery fallback never authorizes source edits. Any accepted candidate still needs the standard implementer/verifier/reviewer/reflector chain.
+- Search breadth is adaptive. Rotate source families and record skipped families; do not require a fixed count of external searches.
+
 ## Deviation Guard (50% Rule)
 
 ### What "50% deviation" means
@@ -161,7 +170,7 @@ Milestones CANNOT be changed without user confirmation IF:
 
 ## Value Reflection (per cycle)
 
-After EVERY spark cycle, CEO writes:
+After EVERY spark cycle, CEO gives the value-reflection bullets and task-scribe writes:
 
 ```text
 ## Cycle N — Value Reflection
@@ -177,6 +186,10 @@ Cumulative deviation (10-cycle): [X]%
 
 If CEO CANNOT write a convincing value reflection → spark was NOT valuable → record as rejected and move on.
 
+### Reflector Escalation
+
+CEO may self-score low-risk candidates during pre-implementation triage, but final acceptance for an implemented spark still requires verifier evidence, cross-review, and reflector PASS. Use a reflector subagent for milestone completion, cumulative deviation concerns, high-risk changes, multi-file implementation, conflicting reviewer findings, or unclear value.
+
 ## Stop Condition
 
 **Spark mode does NOT auto-stop.** Only these events stop it:
@@ -199,6 +212,10 @@ If CEO CANNOT write a convincing value reflection → spark was NOT valuable →
 | SP4 | **Silent North Star drift** | Small changes accumulate, direction shifts without noticing | Cumulative deviation check every 10 cycles |
 | SP5 | **Milestone rot** | Milestones become irrelevant but aren't updated | Milestone review at Re-Anchor Gate |
 | SP6 | **Spark tunnel vision** | Only looking at one type of source | Rotate through the registered source families and record skipped sources |
+| SP7 | **Literal anti-pattern matching** | A rule only catches one exact phrase and misses nearby violations | State the invariant and forbidden behavior, not just a magic string |
+| SP8 | **Checksum drift after doc changes** | Root docs, templates, or `.harness-version` disagree after edits | Sync root + templates, then run `node scripts/build-version.mjs --check` |
+| SP9 | **Process-file token burn** | CEO spends high-reasoning tokens formatting PLAN/PROGRESS/ROADMAP | CEO sends bullets; task-scribe formats task-state writes |
+| SP10 | **Fake delegation** | In-process helper calls are logged as independent workers | Record channel fallback honestly; discovery fallback is not implementation delegation |
 
 ## Integration with /wf-auto
 
@@ -260,7 +277,7 @@ WF-AUTO-SPARK uses a dedicated task capsule at `Harness/tasks/auto/`. Shared wit
 
 ### PLAN.md Format (per cycle)
 
-Before implementing, CEO writes one cycle plan:
+Before implementing, CEO gives one cycle-plan brief and task-scribe writes:
 
 ```markdown
 # Cycle N Plan

package/templates/common/Harness/WF-AUTO.md

@@ -205,7 +205,7 @@ TIER 2 — Did any selected probe find an actionable direction?
   └── NO → Run a confirmation pass with a different scan strategy.
 
 TIER 3 — Is uncertainty still high or coverage borderline?
-  ├── YES → Re-run only the uncertain probe, or invoke the cross-model oracle.
+  ├── YES → Re-run only the uncertain probe, or invoke the peer-review oracle.
   └── NO → Record an empty confirmation pass.
 
 TIER 4 — Two different confirmation strategies are empty?
@@ -215,9 +215,10 @@ TIER 4 — Two different confirmation strategies are empty?
 
 **Oracle Rules (modeled on /wf-review):**
 
-- [ ] CEO detects which CLI is running: `which codex` / `which claude`
-- [ ] CEO invokes the OTHER CLI only when unresolved high-risk uncertainty or borderline coverage justifies it
-- [ ] If neither CLI is available, record "oracle unavailable" in PROGRESS.md and continue with local confirmation
+- [ ] CEO detects available peer CLIs: `claude`, `codex`, and `opencode`
+- [ ] CEO invokes a peer CLI only when unresolved high-risk uncertainty or borderline coverage justifies it
+- [ ] If no peer CLI is available, dispatch the installed `reviewer` role as an independent subagent context
+- [ ] If neither peer CLI nor subagent surface is available, record "oracle unavailable" in PROGRESS.md and continue with local confirmation
 - [ ] Oracle is invoked at most once per adaptive exhaustion event
 - [ ] Oracle findings are treated as severity=high by default (external model perspective gets extra weight)
 
@@ -360,7 +361,7 @@ CEO presents:
 
 ### Spark: External Candidate Provider
 
-Spark is NOT a separate optimization engine. It is a **candidate provider** plugged into W0, alongside the adaptive probe scan and the cross-model oracle. W1 still owns prioritization across ALL sources.
+Spark is NOT a separate optimization engine. It is a **candidate provider** plugged into W0, alongside the adaptive probe scan and the peer-review oracle. W1 still owns prioritization across ALL sources.
 
 **When spark activates:**
 - W0 internal scan returns empty AND oracle also empty → `auto.spark` state
@@ -398,7 +399,7 @@ Spark is NOT a separate optimization engine. It is a **candidate provider** plug
 
 W1 prioritization now handles three source types:
 - `source=internal` — from adaptive probe scan
-- `source=oracle` — from cross-model review
+- `source=oracle` — from peer CLI or reviewer-subagent review
 - `source=spark-<name>` — from external inspiration search
 
 Tie-breaking: internal > oracle > spark (local context beats external inspiration).
@@ -490,14 +491,14 @@ Mini PRD-derived AC IDs in `/wf-auto`.
 |-----------|-----|---------|----------|
 | Scope | Task-bounded | Task-bounded | Unbounded |
 | Stop condition | Task complete | Task complete | Dynamic obligations covered + two different empty confirmation passes |
-| Direction | User-specified | User-specified | AI-inferred + cross-model oracle + external spark |
+| Direction | User-specified | User-specified | AI-inferred + peer-review oracle + external spark |
 | Organization | Flat (CEO + agents) | 3-tier (CEO→Mgr→Worker) | Flat (CEO + selected probes + oracle + spark searchers + build agents) |
 | Duration | One task | One task | Perpetual |
 | User interaction | At key gates | At key gates | Adaptive checkpoint (2→5→10 cycles), 2 questions only |
 | Cycle count | 1 | 1 (multi-wave) | ∞ (until exhaustion) |
 | Files/cycle | Per task | Per wave (many) | ≤3 per cycle |
 | Exploration | 3-5 agents once | 5-10 agents once | Dynamic probes + triggered spark sources per cycle |
-| Cross-model check | No (wf-review is separate) | No (wf-review is separate) | Yes — Cross-Model Oracle built into A-GATE Tier 1.5 |
+| Peer review check | No (wf-review is separate) | No (wf-review is separate) | Yes — peer-review oracle built into A-GATE Tier 1.5 |
 | External inspiration | No | No | Yes — Spark candidate provider when internal sources empty |
 | Evidence tracking | Per task | Per task | Evidence ledger per cycle with measured impact |
 

package/templates/common/Harness/WF-MAX.md

@@ -26,9 +26,43 @@ WF-MAX is **explicit only**. Enter ONLY when the user explicitly types `/wf-max`
 
 CEO reads, plans, dispatches, synthesizes, and writes task state only.
 - **CEO never writes production source code.** All source edits are delegated to Workers.
-- CEO may spawn task-scribe (haiku) to maintain dispatch ledger and heartbeat.
+- CEO spawns task-scribe (haiku) by default to maintain dispatch ledger, heartbeat, and evidence pointers; if unavailable, CEO records the degradation and writes only the smallest durable checkpoint.
 - CEO may spawn codebase-explorer(s) (haiku) for scoped source discovery.
 
+## Worker Channel Degradation & Independence
+
+WF-MAX Workers MUST execute as **independent agent contexts** — never as in-process tool calls from the CEO thread. When a worker channel is unavailable, degrade honestly; never disguise an in-process tool call as a Worker.
+
+### Independence Levels
+
+| Channel | Independence | Counts as Worker? |
+|---|---|---|
+| native subagent (Agent-tool dispatch) | independent | yes |
+| peer CLI (`claude -p`, `codex exec`, `opencode run`) | independent (separate process) | yes |
+| MCP tools (`mcp__codex.codex_implement`, `mcp__claude.claude_implement`, etc.) | inprocess (CEO thread) | NEVER |
+
+### Degradation Chain
+
+When delegating source edits to a Worker, try channels in order; descend on failure; when all independent channels are unavailable, stop honestly and ask the user — do NOT fall back to an in-process MCP tool:
+
+0. **Probe first** — before dispatching any Worker, run `node Harness/scripts/probe-worker-channels.mjs` and read the live matrix (`Harness/tasks/auto/CHANNEL-MATRIX.md`). Degrade based on facts, not assumptions. Start the chain at the highest-priority channel the probe reports `available`.
+1. native subagent (preferred — independent context, bounded writeSet)
+2. `claude -p` peer CLI (independent process)
+3. `codex exec` peer CLI (independent process)
+4. all unavailable → honest pause + human escalation (record blocker in task PROGRESS.md)
+
+### Hard Prohibition
+
+`mcp__codex.codex_implement`, `mcp__claude.claude_implement`, and any CEO-thread MCP tool call MUST NOT be recorded or used as Worker execution. An in-process tool call has no independent context boundary and no enforced writeSet — using it and logging "CEO did not edit source" is fake compliance. Historical instance: `tasks/task-framework-metrics-and-entry-contract/PLAN.md:791`.
+
+### Timeout & Retry
+
+Every channel probe is bounded to ≤15s via `Promise.race` (see `probe-worker-channels.mjs`). A Worker dispatch that returns `unavailable-timeout` or `unavailable-error` (transient) is retried **once**; a second failure descends to the next channel in the chain. WF-MAX MUST NEVER hang on an unresponsive channel (historical failure: a 300s Codex read-only query hang, `tasks/task-framework-metrics-and-entry-contract/PLAN.md`). Bounded probes + single retry + honest descent replace silent hangs.
+
+### Probe the current environment
+
+Run `node Harness/scripts/probe-worker-channels.mjs` to get the live availability + independence matrix (see `tasks/auto/CHANNEL-MATRIX.md`). Degrade based on facts, not assumptions.
+
 ## Parallelism Priority
 
 1. Read-only exploration — max parallel, all readSets
@@ -51,7 +85,7 @@ Manager_max = min(max(Manager_min * 2, Manager_min), 7)
 - One file_claim per write Worker
 - WriteSet must be disjoint across parallel Workers
 - Workers return <=250 tokens + evidence/file paths
-- task-scribe (haiku) runs alongside any wave
+- task-scribe (haiku) runs alongside any wave when available; process-file delegation is the default, not a token-heavy CEO formatting task
 
 ## Token Budget and Fan-Out Caps
 
@@ -80,7 +114,7 @@ D-GATE is mandatory before implementation waves per [WF-KERNEL.md](WF-KERNEL.md)
 
 1. Current runtime subagents first.
 2. Close completed agents; fill idle slots immediately.
-3. Cross-CLI overflow: Codex → `claude -p`, Claude → `codex exec`.
+3. Cross-CLI overflow: use an available peer CLI with explicit dispatch packets: `claude -p`, `codex exec`, or `opencode run --agent <role> --dir .`.
 4. Bounded-pass fallback only when subagents and overflow are exhausted.
 5. Generated Codex config defaults to `agents.max_threads = 12` and `agents.max_depth = 1`. Ask the user before raising `agents.max_threads` above that default. Keep `max_depth = 1` unless recursive delegation is explicitly approved.
 6. Do not rely on Codex++, undocumented config, environment variables, forked/derived conversations, or third-party forks as stable capacity.

package/templates/common/Harness/dispatch.md

@@ -13,7 +13,7 @@ Use when work needs parallel reading, independent review, cross-layer analysis,
 - Important assumptions, decisions, blockers, evidence, and handoffs must be written to `Harness/tasks/<task-id>/PROGRESS.md` and `Harness/tasks/<task-id>/PLAN.md`, the current feature doc, `Harness/MEMORY.md`, or `Harness/memory/*` as appropriate.
 - Resumable state is governed by [WF-STATE.md](WF-STATE.md). On session start, the controller reads STATE.json before building a fresh dispatch table.
 - PRD-derived Acceptance Criteria are the source of truth. Dispatch packets must carry the relevant AC IDs and contracts.
-- Agent count: default (non-WF) <=3 active agents; `/wf` selects a tier dynamically per [WF-KERNEL.md](WF-KERNEL.md) (Light/Standard/Full); `/wf-max` inherits the selected tier and adds maximum safe fan-out (WF-Max-Useful default; WF-Max-Strict explicit only), removing the Harness default cap through the span formula. Real concurrency is still bounded by runtime thread budget, config, billing, and local resources. Use current runtime subagents first, close completed agents, then cross-CLI overflow. Generated Codex config defaults to `agents.max_threads = 12` and `agents.max_depth = 1`; ask the user before raising `agents.max_threads` above that default. See [WF.md](WF.md) and [WF-MAX.md](WF-MAX.md).
+- Agent count: default (non-WF) <=3 active agents; `/wf` selects a tier dynamically per [WF-KERNEL.md](WF-KERNEL.md) (Light/Standard/Full); `/wf-max` inherits the selected tier and adds maximum safe fan-out (WF-Max-Useful default; WF-Max-Strict explicit only), removing the Harness default cap through the span formula. Real concurrency is still bounded by runtime thread budget, config, billing, and local resources. Use current runtime subagents first, close completed agents, then peer-CLI overflow (`claude -p`, `codex exec`, or `opencode run --agent <role> --dir .`). Generated Codex config defaults to `agents.max_threads = 12` and `agents.max_depth = 1`; ask the user before raising `agents.max_threads` above that default. See [WF.md](WF.md) and [WF-MAX.md](WF-MAX.md).
 - Read-only agents may run in parallel.
 - Writing agents run serially unless write sets are disjoint.
 - Use a worktree when two agents may touch overlapping files or long-running branches.

package/templates/common/Harness/scripts/scan-clean.mjs

@@ -23,7 +23,7 @@ import { createInterface } from 'readline';
 const __dirname = dirname(fileURLToPath(import.meta.url));
 const ROOT = process.env.WF_ROOT ? resolve(process.env.WF_ROOT) : resolve(__dirname, '..', '..');
 const VERSION_FILE = resolve(ROOT, 'Harness', '.harness-version');
-const DEFAULT_SOURCE_BASE = 'https://raw.githubusercontent.com/zingspark/create-harness-vibe-coding/main/templates/common/';
+const DEFAULT_SOURCE_BASE = 'https://raw.githubusercontent.com/LiWeny16/create-harness-vibe-coding/main/templates/common/';
 
 // ── Classification constants ────────────────────────────────────────
 
@@ -65,6 +65,7 @@ const MANAGED_SUBDIRS = new Set([
 
 /** Reject paths that escape ROOT (traversal, absolute, .., etc.). */
 function safePath(file) {
+  if (/^[A-Za-z]:/.test(file)) return null;
   let normalized = file.replace(/\\/g, '/').replace(/^\/+/, '');
   if (/\/\//.test(normalized)) return null;
   if (normalized.split('/').some(p => p === '..')) return null;

package/templates/common/Harness/scripts/validate-harness.mjs

@@ -97,6 +97,7 @@ const required = [
   '.opencode/commands/wf-help.md',
   '.opencode/commands/wf-update.md',
   ...opencodeWorkflowCommands.map(command => `.opencode/commands/${command}.md`),
+  '.opencode/plugins/harness-wf-status.mjs',
   '.claude/rules/ecc/common.md',
   ...commonAgents.map(agent => `.claude/agents/${agent}.md`),
   ...commonAgents.map(agent => `.opencode/agents/${agent}.md`),
@@ -324,6 +325,23 @@ for (const taskDir of taskDirs) {
   }
 }
 
+// M2: MCP-as-Worker fake-compliance guard.
+// Task capsules MUST NOT record mcp__codex.codex_implement / mcp__claude.claude_implement
+// as Worker execution. Historical do-not-repeat references are allowed only when the
+// file is marked ANTI-PATTERN. See Harness/WF-MAX.md "Worker Channel Degradation & Independence".
+const MCP_AS_WORKER_RE = /mcp__(?:codex)\.codex_implement|mcp__(?:claude)\.claude_implement/;
+for (const taskDir of taskDirs) {
+  if (!taskDir.startsWith('task-')) continue;
+  for (const f of ['PLAN.md', 'PROGRESS.md']) {
+    const rel = `Harness/tasks/${taskDir}/${f}`;
+    const text = read(rel);
+    if (!text) continue;
+    if (MCP_AS_WORKER_RE.test(text) && !/ANTI-PATTERN/i.test(text)) {
+      errors.push(`${rel} records mcp__*.implement as Worker execution (fake compliance; see WF-MAX.md "Worker Channel Degradation & Independence"). Mark historical references with "ANTI-PATTERN" or remove the tool call.`);
+    }
+  }
+}
+
 if (fs.existsSync(path.join(root, 'Harness/research/scaffolds.md'))) {
   errors.push('legacy research file should be renamed: Harness/research/scaffolds.md -> Harness/research/research-results.md');
 }
@@ -710,6 +728,12 @@ requireText('Harness/WF-AUTO.md', 'reflector PASS', 'wf-auto reflector gate');
 requireText('Harness/WF-AUTO-SPARK.md', 'Inherited Execution Chain', 'wf-auto-spark inherited execution chain');
 requireText('Harness/WF-AUTO-SPARK.md', 'External spark search replaces discovery only', 'wf-auto-spark discovery-only inheritance');
 requireText('Harness/WF-AUTO-SPARK.md', 'reflector PASS', 'wf-auto-spark reflector gate');
+requireText('Harness/WF-AUTO-SPARK.md', 'task-scribe formats task-state writes', 'wf-auto-spark task-scribe recorder delegation');
+requireText('Harness/WF-AUTO-SPARK.md', 'searchFallback: ceo-direct', 'wf-auto-spark search fallback');
+requireText('Harness/WF-AUTO-SPARK.md', 'pre-implementation triage', 'wf-auto-spark reflector escalation');
+requireText('Harness/WF-AUTO-SPARK.md', 'Literal anti-pattern matching', 'wf-auto-spark anti-pattern invariant guard');
+requireText('Harness/WF-AUTO-SPARK.md', 'node scripts/build-version.mjs --check', 'wf-auto-spark checksum drift guard');
+requireText('Harness/WF-MAX.md', 'process-file delegation is the default', 'WF-MAX task-scribe process-file delegation default');
 requireText('Harness/MEMORY_PROTOCOL.md', 'Scenario Memory Hints', 'memory protocol scenario hints');
 requireText('Harness/MEMORY_PROTOCOL.md', 'WF closeout', 'memory protocol WF closeout injection row');
 requireText('Harness/MEMORY_PROTOCOL.md', 'memory-master owns writes', 'memory protocol memory write ownership');
@@ -725,6 +749,11 @@ requireText('.claude/skills/tdd/SKILL.md', 'No syntax-only acceptance', 'tdd ski
 requireText('.claude/skills/wf-remove/SKILL.md', 'User-facing removal is the slash/skill command', 'wf-remove slash command is user-facing');
 requireText('.claude/skills/wf-remove/SKILL.md', 'agent-internal execution steps', 'wf-remove script commands are agent-internal');
 requireText('.claude/skills/wf-remove/SKILL.md', 'verify residual discovery folders', 'wf-remove residual discovery verification');
+requireText('.claude/skills/wf-review/SKILL.md', 'opencode run --agent reviewer', 'wf-review OpenCode peer CLI path');
+requireText('.claude/skills/wf-review/SKILL.md', 'Role: reviewer', 'wf-review installed reviewer role fallback');
+requireText('.claude/skills/wf-review/SKILL.md', 'The main agent is the controller', 'wf-review controller final authority');
+requireText('.opencode/commands/wf-review.md', 'peer-review contract', 'OpenCode wf-review wrapper peer-review contract');
+requireText('Harness/subagents.md', 'For `/wf-review`, use the installed `reviewer` role', 'subagents wf-review role fallback');
 requireText('.claude/agents/tdd-guide.md', 'Browser Acceptance Rules', 'tdd-guide browser acceptance rules');
 requireText('.claude/agents/tdd-guide.md', 'real user actions', 'tdd-guide real user action requirement');
 requireText('.claude/agents/test-writer.md', 'Harness/ACCEPTANCE_PROTOCOL.md', 'test-writer loads acceptance protocol');
@@ -766,6 +795,12 @@ forbidText('.claude/settings.json', 'wf-mode-hook.mjs', 'Claude WF hook command
 forbidText('.codex/hooks.json', 'wf-mode-hook.mjs', 'Codex WF hook command registration');
 const codexHookConfig = read('.codex/hooks.json');
 const claudeSettings = read('.claude/settings.json');
+requireText('.codex/hooks.json', '"SessionStart"', 'Codex startup-only update hook');
+requireText('.claude/settings.json', '"SessionStart"', 'Claude startup-only update hook');
+forbidText('.codex/hooks.json', 'UserPromptSubmit', 'Codex turn-by-turn update hook');
+forbidText('.claude/settings.json', 'UserPromptSubmit', 'Claude turn-by-turn update hook');
+requireText('.opencode/plugins/harness-wf-status.mjs', 'opencode.startup', 'OpenCode startup-only update check');
+forbidText('.opencode/plugins/harness-wf-status.mjs', "'chat.message'", 'OpenCode turn-by-turn update hook');
 if (codexHookConfig && !codexHookConfig.includes('wf-auto')) {
   errors.push('.codex/hooks.json may only exist for a wf-auto hook configuration');
 }
@@ -840,6 +875,7 @@ requireText('Harness/WF-KERNEL.md', 'small-fast', 'WF-KERNEL small-fast tier map
 // WF variants reference WF-KERNEL
 requireText('Harness/WF.md', 'WF-KERNEL.md', 'WF.md references WF-KERNEL');
 requireText('Harness/WF-MAX.md', 'WF-KERNEL.md', 'WF-MAX.md references WF-KERNEL');
+requireText('Harness/WF-MAX.md', 'probe-worker-channels.mjs', 'WF-MAX references worker-channel probe (M3 anti-regression)');
 
 // Tier-aware acceptance: WF-Light must NOT require global cross-review/reflector
 requireText('Harness/WF.md', 'Cross-review and reflector NOT mandatory', 'WF.md WF-Light no mandatory cross-review');

package/templates/common/Harness/scripts/wf-auto-update-prompt.mjs

@@ -142,7 +142,7 @@ function buildMessage(update) {
     `Harness update available: ${from} -> ${to}.`,
     `Before unrelated work, ask the user whether to run /wf-update.`,
   ];
-  // Only include summary counts — never list individual conflict files in hook output
+  // Only include summary counts; never list individual conflict files in hook output.
   if (update.conflict > 0) {
     parts.push(`${update.conflict} conflict file(s) will need agent/user merge decisions.`);
   }
@@ -174,7 +174,7 @@ function emit(message, payload = {}) {
     if (context) {
       console.log(JSON.stringify({
         hookSpecificOutput: {
-          hookEventName: input.hook_event_name || 'UserPromptSubmit',
+          hookEventName: input.hook_event_name || 'SessionStart',
           additionalContext: context,
         },
       }));

package/templates/common/Harness/scripts/wf-remove.mjs

@@ -214,6 +214,7 @@ const CLEANUP_DIRS = [
 
 /** Reject paths that escape ROOT (traversal, absolute, .., etc.). Sync with wf-update-check. */
 function safePath(file) {
+  if (/^[A-Za-z]:/.test(file)) return null;
   let normalized = file.replace(/\\/g, '/').replace(/^\/+/, '');
   if (normalized.includes('//')) return null;  // double slash bypass
   if (normalized.split('/').some(p => p === '..')) return null;