wogiflow 2.22.4 → 2.24.0

package/.claude/commands/wogi-debug-hypothesis.md

@@ -2,26 +2,84 @@
 description: "Parallel hypothesis debugging - spawns agents for competing theories"
 effort: high
 ---
-Parallel hypothesis debugging - spawns multiple agents to investigate competing theories simultaneously.
+Parallel hypothesis debugging — spawns multiple agents to investigate competing theories simultaneously. **v2.23.0+**: now runs IGR-style assumption surfacing and scope verification BEFORE generating hypotheses, and a cross-hypothesis adversary step AFTER investigation to catch "three variations of the same wrong theory."
 
 ## Usage
 
 ```
 /wogi-debug-hypothesis "description of the bug or unexpected behavior"
+/wogi-debug-hypothesis --no-assumptions   # skip Tier 2 assumption gate (not recommended)
+/wogi-debug-hypothesis --no-adversary     # skip cross-hypothesis adversary
 ```
 
-## How It Works
+## How It Works (v2.23.0+)
 
-1. **Analyze** the bug description to generate 2-3 competing hypotheses
-2. **Spawn** parallel Task agents, each investigating one hypothesis
-3. **Consolidate** findings into a single diagnosis
-4. **Recommend** the most likely root cause with evidence
+```
+Bug description
+  ↓
+Step 0: Scope-Confidence pre-check — verify files/modules you plan to
+        investigate actually exist (regex → grep); halt if contradictions
+  ↓
+Step 1: Assumption Surfacing (Tier 2) — list domain-model assumptions
+        your hypotheses will depend on; WAIT for user confirmation
+  ↓
+Step 2: Generate 2-3 hypotheses grounded in CONFIRMED assumptions
+  ↓
+Step 3: Spawn parallel Explore agents per hypothesis (READ-ONLY)
+  ↓
+Step 4: Consolidate findings
+  ↓
+Step 5: Hypothesis Adversary — spawn adversary on different model; goal:
+        challenge the winning hypothesis against rejected ones; surface
+        any overlap that suggests the diagnosis missed the real cause
+  ↓
+Step 6: Final diagnosis with confidence + suggested fix
+```
+
+The "user is the adversary" assumption pass (Step 1) is the single most
+important addition — it forces hypothesis generation to be grounded in
+facts rather than whatever the AI's first read of the bug happened to be.
 
 ## Execution Steps
 
-### Step 1: Generate Hypotheses
+### Step 0: Scope-Confidence Pre-Check (v2.23.0+)
+
+Before generating hypotheses, extract noun-phrases from the bug description that look like code entities (function names, filenames, module names, API endpoints) and verify they exist in the codebase:
+
+```javascript
+const gates = require('wogiflow/scripts/flow-story-gates');
+const audit = gates.auditScopeConfidence(ARGUMENTS);
+```
+
+If any assumption is **CONTRADICTED** (user refers to a component that doesn't exist), HALT and ask: "You mentioned `<X>` but I don't see it in the codebase. Did you mean `<nearest match>`, or is `<X>` created under a different name?"
+
+This catches the "investigate imaginary code" failure mode before wasting a parallel agent run.
+
+### Step 1: Assumption Surfacing (Tier 2 — MANDATORY unless `--no-assumptions`)
+
+Before generating ANY hypothesis, identify the domain-model assumptions your theories will depend on. Present them in a fenced block and **WAIT** for user confirmation.
+
+```
+━━━ ASSUMPTIONS (confirm before I generate hypotheses) ━━━
+The bug description is: "[ARGUMENTS]"
+
+My hypothesis generation will assume:
+  1. <assumption about what X does>
+  2. <assumption about when Y fires>
+  3. <assumption about data shape>
+  4. <assumption about call ordering>
+
+Do these match your understanding? [confirm / correct <N>]
+━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
+```
+
+**Do NOT proceed to Step 2 while waiting.** The user's response becomes the ground truth for hypothesis generation.
+
+**Rationale** (from `config.researchReasoningGate`): same-model self-critique rubber-stamps. The USER is the effective adversary — they validate the domain model before the AI builds theories on invisible guesses. Three hypotheses rooted in the same wrong assumption are still all wrong.
 
-Read the bug description from ARGUMENTS and generate 2-3 hypotheses.
+### Step 2: Generate Hypotheses
+
+Using the confirmed assumptions from Step 1, generate 2-3 hypotheses.
 
 For each hypothesis, identify:
 - **Theory**: What might be causing this
@@ -52,7 +110,7 @@ Spawning 3 investigation agents in parallel...
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 ```
 
-### Step 2: Spawn Parallel Investigators
+### Step 3: Spawn Parallel Investigators
 
 Launch one Task agent per hypothesis. **All agents must be launched in a single message** (parallel Task calls).
 
@@ -91,7 +149,7 @@ IMPORTANT: Only use read-only tools (Glob, Grep, Read, WebSearch, WebFetch). Do
 
 Use `subagent_type=Explore` for all investigation agents.
 
-### Step 3: Consolidate Findings
+### Step 3.5: Consolidate Findings
 
 After all agents complete, display the consolidated results:
 
@@ -116,7 +174,61 @@ After all agents complete, display the consolidated results:
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 ```
 
-### Step 4: Diagnosis
+### Step 4: Hypothesis Adversary (v2.23.0+ — MANDATORY unless `--no-adversary`)
+
+After consolidation, spawn a single Agent (different `model` param if `config.hybrid.enabled`, else same) with this prompt:
+
+```
+You are the hypothesis adversary.
+
+Confirmed domain assumptions (from the user):
+  [list from Step 1]
+
+The 3 hypotheses and their verdicts:
+  H1 [CONFIRMED, confidence HIGH]: <theory>
+    Evidence: [top 3 findings]
+  H2 [REFUTED, confidence HIGH]: <theory>
+    Evidence: ...
+  H3 [INCONCLUSIVE, confidence LOW]: <theory>
+    Evidence: ...
+
+Winning hypothesis: H1
+
+Your job (10 min):
+  1. Does H1's evidence actually prove H1, or could it also be consistent
+     with H2 or H3? List overlapping evidence.
+  2. Is there a 4th hypothesis that would explain ALL the evidence better
+     than H1 alone?
+  3. List 1-3 specific reasons H1 might be WRONG despite the verdict.
+  4. Cite file:line for each concern.
+
+Output format:
+{
+  "overlap_risk": "low|medium|high",
+  "alternative_hypothesis": "<new theory or 'none'>",
+  "concerns": [
+    { "concern": "...", "evidence": "<file:line>" }
+  ]
+}
+```
+
+Present adversary critique to the user alongside the original diagnosis:
+
+```
+━━━ DIAGNOSIS ━━━
+[original Step 5 content]
+
+━━━ ADVERSARY CRITIQUE (reviewed by a different model) ━━━
+Overlap risk: [low|medium|high]
+Alternative hypothesis: [new theory or "none"]
+Concerns:
+  • [concern 1] — [file:line]
+  • [concern 2] — [file:line]
+```
+
+One pass only — no iteration loop (this is debug, not implementation).
+
+### Step 5: Diagnosis
 
 Synthesize the findings into a final diagnosis:
 

package/.claude/commands/wogi-epics.md

@@ -165,6 +165,18 @@ You MAY:
 
 **If the user provides 9 items, the epic MUST contain 9 stories (or items grouped into stories where every item appears as an acceptance criterion). Verify this with a reconciliation count before proceeding.**
 
+## P0 Specification-Quality Gates (v2.24.0+)
+
+Before finalizing an epic, run the same P0 gates `/wogi-story` uses (`scripts/flow-story-gates.js`):
+
+1. **Long Input Gate** — ≥40 lines or ≥5 items → route to `/wogi-extract-review`
+2. **Item Reconciliation** — already enforced via Anti-Deferral above; use `gates.reconcileItems()` to generate the manifest mechanically
+3. **Consumer Impact** — if epic description contains refactor/rename/migrate/etc. → grep consumers, list breaking count, recommend phased migration at ≥5
+4. **Scope-Confidence** — extract "new X"/"existing Y" assumptions → classify against codebase → surface CONTRADICTED/UNVERIFIED as Pending Clarifications
+5. **Intent Bootstrap Coordination** — schedule IGR bootstrap if missing; respects existing session-state flag
+
+All fail-open. Call `require('wogiflow/scripts/flow-story-gates')` directly in the epic-creation flow.
+
 ## Tips
 
 - **Start with epics for major features** - Break down into stories before implementation

package/.claude/commands/wogi-extract-review.md

@@ -182,6 +182,31 @@ Contradictions are resolved automatically using temporal ordering:
 
 The superseded statement is marked and excluded from story generation.
 
+## Item Manifest Export (v2.24.0+)
+
+After review is complete, the confirmed items can be exported as an **Item Manifest** in the format `/wogi-story`'s P0 item-reconciliation gate consumes directly. This closes the re-extraction loop — when `/wogi-start` routes a long input to `/wogi-extract-review`, the resulting manifest is handed to `/wogi-story` with `bypassLongInput: true` so story creation doesn't re-route back to extraction.
+
+```bash
+flow extract-zero-loss manifest    # JSON: { items, count, bypassLongInput, sourceSessionId, intentBootstrapScheduled }
+```
+
+AI flow:
+```
+User pastes large input
+  ↓
+/wogi-start detects line/item threshold → routes to /wogi-extract-review
+  ↓
+Extract → review → confirm completeness
+  ↓
+flow extract-zero-loss manifest → JSON
+  ↓
+/wogi-story --full-input=<manifest.items.join('\n')> --bypass-long-input
+  ↓
+Item Reconciliation gate sees bypassLongInput, skips Gate 1, runs Gates 2-5 normally
+```
+
+The manifest also carries an `intentBootstrapScheduled` flag so `/wogi-story` won't re-prompt the user if /wogi-extract-review already scheduled IGR bootstrap.
+
 ## Files
 
 | File | Location |

package/.claude/commands/wogi-feature.md

@@ -145,6 +145,41 @@ node node_modules/wogiflow/scripts/flow-feature.js progress ft-a1b2c3d4
 | → | In Progress (1-99%) |
 | ✓ | Completed (100%) |
 
+## Anti-Deferral Rule (MANDATORY — v2.24.0+)
+
+**When creating a feature from user input, EVERY item the user provided MUST become a tracked story within the feature.**
+
+You must NEVER:
+- Create stories for items 1-3 and silently skip items 4-6 because you judged them as "enhancements"
+- Label items as "deferred" or "long-term" and exclude them from the feature
+- Apply your own priority filter to decide which items deserve stories
+
+You MAY:
+- Assign different priorities (P0/P1/P2/P3) — but ALL items get stories
+- Suggest an execution order — but ALL items are tracked in the feature
+- Ask the user "Should I defer items 4-6?" — explicit user consent is the ONLY valid reason to exclude items
+
+**If the user provides 5 items, the feature MUST contain 5 stories (or items grouped into stories where every item appears as an acceptance criterion).** Verify with a reconciliation count before proceeding.
+
+## P0 Specification-Quality Gates (v2.24.0+)
+
+When creating a feature from user input, apply the same P0 gates `/wogi-story` uses (from `scripts/flow-story-gates.js`):
+
+1. **Long Input Gate** — ≥40 lines or ≥5 items → route to `/wogi-extract-review` first
+2. **Item Reconciliation** — ≥3 items → enumerate items as a manifest + verify every item maps to at least one story
+3. **Consumer Impact Analysis** — if the feature title/description contains refactor/rename/migrate/replace/consolidate/split/extract/move keywords → run `git grep` on seeded filenames, list likely consumers, recommend phased migration if ≥5 breaking consumers
+4. **Scope-Confidence Audit** — extract "new X"/"existing Y"/"the Z service" assumptions from the description, classify against codebase (VERIFIED/CONTRADICTED/UNVERIFIED), write to a Pending Clarifications section on the feature
+5. **Intent Bootstrap Coordination** — if IGR is enabled and artifacts missing, schedule bootstrap via session-state flag (do NOT duplicate-prompt)
+
+All gates fail-open. Use `flow-story-gates.js` directly:
+
+```javascript
+const gates = require('wogiflow/scripts/flow-story-gates');
+gates.reconcileItems(userInput);
+gates.analyzeConsumerImpact(title + ' ' + description);
+gates.auditScopeConfidence(description);
+```
+
 ## Tips
 
 - **Features represent user-facing capabilities** - Not technical components

package/.claude/commands/wogi-morning.md

@@ -30,6 +30,20 @@ KEY CONTEXT
 BLOCKERS
   - Waiting on backend team for /preferences endpoint
 
+WORKSPACE DISPATCH ISSUES (v2.23.0 — manager mode only)
+  ━━━ OVERDUE WORKSPACE DISPATCHES (2) ━━━
+  • wf-a1b2c3d4 → worker-1  | dispatched 47m ago (17m past 30m deadline)
+  • wf-e5f67890 → worker-2  | dispatched 2h3m ago (1h33m past 30m deadline)
+
+  ━━━ LOST DISPATCHES — WORKER RESTARTED WITH EMPTY QUEUE (1) ━━━
+  • wf-55555555 → worker-3  | dispatched 4m ago | still pending after worker restart
+
+HONESTY CHECK — claim/state contradictions (v2.23.0)
+  ⚠ 2 tasks with claim-vs-state mismatch
+    wf-abc12345 (class A, field notes): "fully completed, shipped yesterday..."
+    wf-def67890 (class B, field result): "0 outages reported, clean rollout..."
+  Review: flow health
+
 RULE VIOLATIONS (deferred from rule creation)
   Rule: "Split Prisma schema into domain files" (added 2026-02-20)
   Violations: 4 across 1 file

package/.claude/commands/wogi-peer-review.md

@@ -39,30 +39,40 @@ Models are selected once per session and remembered for subsequent runs.
 
 **Config**: Models configured in `.workflow/config.json` under `models.providers`. API keys in `.env`.
 
-## Review Flow
+## Review Flow (v2.23.0+)
 
 ```
 ┌─────────────────────────────────────────────────────────┐
 │  /wogi-peer-review                                       │
 ├─────────────────────────────────────────────────────────┤
 │  1. Collect code changes (git diff or specified files)   │
-│  2. Generate improvement-focused prompt                  │
-│  3. If includeClaude enabled:                            │
+│  2. Classify change size → effort tier:                  │
+│     L0/L1 (>10 files)  → opus-4-7 xhigh                  │
+│     L2 (3-10 files)    → sonnet medium                   │
+│     L3 (<3 files)      → haiku medium                    │
+│  3. Generate improvement-focused prompt                  │
+│  4. If includeClaude enabled:                            │
 │     - Launch Claude review (Task agent, Explore type)    │
-│  4. External model(s) review via API                     │
-│  5. Collect all results                                  │
-│  6. Compare findings:                                    │
+│  5. External model(s) review via API                     │
+│  6. Collect all results, tag each claim with Evidence    │
+│     Tier 0-4 (NONE/STATIC/COMPILED/INTERACTIVE/SHIPPED)  │
+│  7. Compare findings:                                    │
 │     - All agree → Strong suggestion                      │
 │     - Partial agree → Present perspectives               │
 │     - Disagree → Surface disagreement                    │
-│  7. Claude synthesizes and responds to feedback           │
-│  8. Output final synthesis                               │
+│  8. Synthesis Adversary (v2.23.0 — NEW):                 │
+│     spawn cross-model agent on DIFFERENT model to        │
+│     critique the synthesis itself — does "3/3 agreement" │
+│     actually mean "3 models all hallucinated the same    │
+│     thing"?                                              │
+│  9. Claude synthesizes + incorporates adversary critique │
+│  10. Output final synthesis with evidence-tier per claim │
 └─────────────────────────────────────────────────────────┘
 ```
 
 ## Review Prompt Template
 
-The peer review prompt focuses on improvements, not correctness:
+The peer review prompt focuses on improvements, not correctness, and now demands evidence tiers per claim (v2.23.0+):
 
 ```
 Review this code for IMPROVEMENT OPPORTUNITIES, not bugs:
@@ -73,9 +83,47 @@ Review this code for IMPROVEMENT OPPORTUNITIES, not bugs:
 4. **Readability**: Could this be clearer/simpler?
 5. **Extensibility**: Will this be easy to extend?
 
-Respond with: specific suggestions, alternative approaches, trade-off analysis.
+For EACH suggestion, tag it with an evidence tier:
+  Tier 0 (NONE)         — no evidence, pure opinion
+  Tier 1 (STATIC)       — based on reading the code
+  Tier 2 (COMPILED)     — would affect compile / type errors
+  Tier 3 (INTERACTIVE)  — would affect runtime behavior observably
+  Tier 4 (SHIPPED)      — you can cite a production incident
+
+Respond with: specific suggestions, alternative approaches, trade-off
+analysis, EACH carrying an explicit evidence tier.
 ```
 
+## Synthesis Adversary (v2.23.0+ — MANDATORY unless `--no-adversary`)
+
+After initial synthesis, spawn a single adversary agent on a DIFFERENT model from the synthesizer (default: if synthesizer is Opus, adversary is Sonnet; config via `peerReview.adversaryModel`). Prompt:
+
+```
+You are the synthesis adversary.
+
+Synthesis claims:
+  • [claim 1, tier 2, agreed by 3 models]
+  • [claim 2, tier 1, agreed by 2 models]
+  • [claim 3, tier 0, 1-model unique insight]
+
+Your job (5 min):
+  1. For each claim: could "agreement" be shared hallucination? What would
+     refute the claim that couldn't have been seen by all 3 reviewers?
+  2. For Tier 0/1 claims: is the evidence actually there in the code, or
+     is it vibes-based?
+  3. Is there an important suggestion MISSING from the synthesis that a
+     human reviewer would flag?
+
+Output:
+{
+  "shared_hallucination_risk": [list of claim indices with reason],
+  "vibes_based_claims": [list of claim indices],
+  "missing_suggestions": [list of suggestions the synthesis missed]
+}
+```
+
+Merge adversary output into the final report — downgrade any claim flagged as shared-hallucination or vibes-based by one evidence tier.
+
 ## Output
 
 ```
@@ -150,5 +198,8 @@ For manual review (no API keys needed): `/wogi-peer-review --manual`
 | `--json` | Output JSON for automation |
 | `--verbose` | Show full model responses |
 | `--create-tasks` | Auto-create tasks for strong agreements |
+| `--no-adversary` | Skip the v2.23.0 synthesis adversary (not recommended for L0/L1 diffs) |
+| `--adversary-model <id>` | Override adversary model (default: cross-model from synthesizer) |
+| `--effort <level>` | Override effort tier (low/medium/high/xhigh/max) — otherwise derived from diff size |
 
 ARGUMENTS: {args}

package/.claude/commands/wogi-plan.md

@@ -182,6 +182,34 @@ When `/wogi-plan` is invoked with a description argument, it should:
 1. Create the plan structure in `.workflow/plans/`
 2. Enter Claude Code plan mode with the description: `EnterPlanMode` with the plan context
 
+## Anti-Deferral Rule (MANDATORY — v2.24.0+)
+
+**When creating a plan from user input, EVERY item the user provided MUST become a tracked epic or feature within the plan.**
+
+You must NEVER:
+- Create epics for items 1-3 and silently skip items 4-7 because you judged them as "enhancements" or "long-term"
+- Label items as "deferred" and exclude them from the plan
+- Apply your own filter to decide which items deserve tracking
+
+You MAY:
+- Assign different priorities (P0/P1/P2/P3) — but ALL items get epics/features
+- Suggest an execution order — but ALL items are tracked in the plan
+- Ask the user "Should I defer items 4-7?" — explicit user consent is the ONLY valid reason to exclude items
+
+**If the user provides 7 items, the plan MUST contain 7 tracked sub-items (epics or features, possibly grouped where every item appears as a criterion).** Verify with a reconciliation count before proceeding.
+
+## P0 Specification-Quality Gates (v2.24.0+)
+
+When creating a plan from user input, apply the same P0 gates `/wogi-story` uses (`scripts/flow-story-gates.js`):
+
+1. **Long Input Gate** — ≥40 lines or ≥5 items → route to `/wogi-extract-review`
+2. **Item Reconciliation** — ≥3 items → enumerate manifest + verify every item maps
+3. **Consumer Impact** — refactor keywords trigger a grep; ≥5 breaking consumers → recommend phased migration
+4. **Scope-Confidence** — extract "new X"/"existing Y"/"the Z service" claims; classify against codebase; surface contradictions as Pending Clarifications
+5. **Intent Bootstrap Coordination** — schedule IGR bootstrap if missing (don't re-prompt)
+
+All fail-open.
+
 ## Tips
 
 - **Plans are for strategic visibility** - Track high-level progress

package/.claude/commands/wogi-session-end.md

@@ -10,8 +10,10 @@ Steps:
 3. **Check app-map** - If new components created, verify they're added
 4. **Update progress.md** - Add handoff notes for next session
 5. **Community push** - If `config.community.enabled`, push anonymous learnings
-6. **Commit changes** - Stage and commit all workflow files
-7. **Offer to push** - Ask if should push to remote
+6. **Completion-claim honesty scan** - Surface done-in-text-but-not-in-status contradictions (2026-04-16 honesty-infra)
+7. **Workspace session-end message (v2.23.0+)** - If running inside a workspace manager, write a `heads-up` message to `.workspace/messages/` so workers know no new dispatches are coming
+8. **Commit changes** - Stage and commit all workflow files
+9. **Offer to push** - Ask if should push to remote
 
 Output:
 ```

package/.claude/commands/wogi-test-browser.md

@@ -328,3 +328,14 @@ Controlled by `.workflow/config.json`:
 - **WebMCP assertions**: ~100 tokens per step (tool call + JSON assertion)
 - **Savings**: ~95% token reduction for a 10-step test flow
 - **Bonus**: Deterministic results (no visual diff ambiguity)
+
+## Evidence Tier (v2.24.0+)
+
+A successful `/wogi-test-browser` run emits **Evidence Tier 4 (SHIPPED)** — the highest tier on the Completion Truth Gate scale. Browser-driven WebMCP tests exercise real DOM interactions with real event listeners, network calls, and state transitions.
+
+When reporting results:
+```json
+{ "evidenceTier": 4, "evidenceTierLabel": "SHIPPED" }
+```
+
+See `/wogi-test` for the full tier scale. L0/L1 tasks touching UI should reach Tier 4 before closing; Truth Gate (Step 3.9) will downgrade "done" claims that don't.

package/.claude/commands/wogi-test-generate.md

@@ -99,4 +99,15 @@ During `/wogi-start` Step 3, verify:
 - Run generated tests AFTER implementing → they should all pass
 - If any test passes before implementation → WARNING: test may be trivial
 
+## Evidence Tier (v2.24.0+)
+
+Generated tests that only assert structural properties (class exists, function has signature) emit **Evidence Tier 1 (STATIC)**. Tests that exercise actual behavior emit **Evidence Tier 2 (COMPILED)** once they pass. To reach Tier 3+ (interactive / shipped), tests must make real network or DOM calls — see `/wogi-test` (API/UI) and `/wogi-test-browser` for those.
+
+When recording test results in `.workflow/verifications/`, include:
+```json
+{ "evidenceTier": 1|2, "evidenceTierLabel": "STATIC"|"COMPILED" }
+```
+
+The Completion Truth Gate reads these labels to decide whether "tests pass" is sufficient to accept a "done" claim. L0/L1 tasks cannot close on Tier 1 alone.
+
 ARGUMENTS: {args}

package/.claude/commands/wogi-test.md

@@ -285,6 +285,31 @@ Include results in the summary:
 Generated Tests: [passed]/[total] passed
 ```
 
+## Evidence Tiers (v2.24.0+)
+
+Every test invocation should emit an **evidence tier** label that the Completion Truth Gate (Step 3.9 in `/wogi-start`) uses to accept or downgrade "done" claims. A task that claims completion without sufficient tier evidence gets surfaced as "implemented (unverified)" rather than rubber-stamped as done.
+
+| Tier | Label | What it means | Source |
+|------|-------|---------------|--------|
+| 0 | NONE | No test ran or all tests silently skipped | N/A |
+| 1 | STATIC | Code parses / type-checks cleanly | `flow lint`, `flow typecheck` |
+| 2 | COMPILED | Unit tests pass | `flow-test-integrity.js` generated/unit tests |
+| 3 | INTERACTIVE | API calls succeed end-to-end (HTTP round-trips, DB reads) | `flow-test-api.js`, `flow-test-ui.js` backend hits |
+| 4 | SHIPPED | UI interaction succeeds in a real browser (click, submit, see result) | `flow-test-ui.js` browser E2E |
+
+Output format (JSON mode):
+```json
+{
+  "passed": true,
+  "results": [...],
+  "evidenceTier": 3,
+  "evidenceTierLabel": "INTERACTIVE",
+  "gates": { "static": true, "unit": true, "api": true, "ui": false }
+}
+```
+
+Any `/wogi-start` task closing with `evidenceTier < 3` for features that touch UI or a service boundary will be flagged by the Truth Gate. L3 tasks (refactor/chore) accept tier 1–2. L0/L1 tasks require tier 3+.
+
 ## Important Notes
 
 - Testing is **disabled by default** — zero overhead for projects that don't use it

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wogiflow",
-  "version": "2.22.4",
+  "version": "2.24.0",
   "description": "AI-powered development workflow management system with multi-model support",
   "main": "lib/index.js",
   "bin": {
@@ -10,7 +10,7 @@
   },
   "scripts": {
     "flow": "./scripts/flow",
-    "test": "NODE_ENV=test node --test tests/auto-compact-prompt.test.js tests/flow-paths.test.js tests/flow-io.test.js tests/flow-config-loader.test.js tests/flow-damage-control.test.js tests/flow-output.test.js tests/flow-constants.test.js tests/flow-session-state.test.js tests/flow-hooks-integration.test.js tests/flow-utils.test.js tests/flow-security.test.js tests/flow-memory-db.test.js tests/flow-durable-session.test.js tests/flow-skill-matcher.test.js tests/flow-bridge.test.js tests/flow-proactive-compact.test.js tests/flow-cascade-completion.test.js tests/flow-capture-gate.test.js tests/flow-correction-detector-hybrid.test.js tests/flow-promote.test.js tests/flow-archive-runs.test.js tests/flow-memory.test.js tests/flow-hooks-pre-tool-helpers.test.js tests/flow-hooks-bugfix-scope-gate.test.js tests/flow-hooks-routing-gate.test.js tests/flow-hooks-phase-read-gate.test.js tests/flow-hooks-commit-log-gate.test.js tests/flow-hooks-deploy-gate.test.js tests/flow-hooks-todowrite-gate.test.js tests/flow-hooks-git-safety-gate.test.js tests/flow-hooks-scope-mutation-gate.test.js tests/flow-hooks-strike-gate.test.js tests/flow-hooks-component-check.test.js tests/flow-hooks-scope-gate.test.js tests/flow-hooks-implementation-gate.test.js tests/flow-hooks-research-gate.test.js tests/flow-hooks-loop-check.test.js tests/flow-hooks-manager-boundary-gate.test.js tests/flow-hooks-phase-gate.test.js tests/flow-hooks-pre-tool-orchestrator.test.js tests/flow-hooks-observation-capture.test.js tests/flow-hooks-task-gate.test.js tests/flow-durable-session-suspension.test.js tests/flow-health-mcp-scopes.test.js tests/flow-lean-config.test.js tests/flow-workspace-autopickup.test.js tests/flow-worker-boundary-gate.test.js tests/flow-worker-question-classifier.test.js tests/flow-completion-truth-gate-contradictions.test.js tests/flow-structure-sensor.test.js tests/flow-workspace-dispatch-tracking.test.js tests/flow-story-gates.test.js tests/flow-workspace-restart-handoff.test.js tests/flow-wogi-claude-wrapper.test.js && NODE_ENV=test node tests/run-quality-gates.test.js",
+    "test": "NODE_ENV=test node --test tests/auto-compact-prompt.test.js tests/flow-paths.test.js tests/flow-io.test.js tests/flow-config-loader.test.js tests/flow-damage-control.test.js tests/flow-output.test.js tests/flow-constants.test.js tests/flow-session-state.test.js tests/flow-hooks-integration.test.js tests/flow-utils.test.js tests/flow-security.test.js tests/flow-memory-db.test.js tests/flow-durable-session.test.js tests/flow-skill-matcher.test.js tests/flow-bridge.test.js tests/flow-proactive-compact.test.js tests/flow-cascade-completion.test.js tests/flow-capture-gate.test.js tests/flow-correction-detector-hybrid.test.js tests/flow-promote.test.js tests/flow-archive-runs.test.js tests/flow-memory.test.js tests/flow-hooks-pre-tool-helpers.test.js tests/flow-hooks-bugfix-scope-gate.test.js tests/flow-hooks-routing-gate.test.js tests/flow-hooks-phase-read-gate.test.js tests/flow-hooks-commit-log-gate.test.js tests/flow-hooks-deploy-gate.test.js tests/flow-hooks-todowrite-gate.test.js tests/flow-hooks-git-safety-gate.test.js tests/flow-hooks-scope-mutation-gate.test.js tests/flow-hooks-strike-gate.test.js tests/flow-hooks-component-check.test.js tests/flow-hooks-scope-gate.test.js tests/flow-hooks-implementation-gate.test.js tests/flow-hooks-research-gate.test.js tests/flow-hooks-loop-check.test.js tests/flow-hooks-manager-boundary-gate.test.js tests/flow-hooks-phase-gate.test.js tests/flow-hooks-pre-tool-orchestrator.test.js tests/flow-hooks-observation-capture.test.js tests/flow-hooks-task-gate.test.js tests/flow-durable-session-suspension.test.js tests/flow-health-mcp-scopes.test.js tests/flow-lean-config.test.js tests/flow-workspace-autopickup.test.js tests/flow-worker-boundary-gate.test.js tests/flow-worker-question-classifier.test.js tests/flow-completion-truth-gate-contradictions.test.js tests/flow-structure-sensor.test.js tests/flow-workspace-dispatch-tracking.test.js tests/flow-story-gates.test.js tests/flow-workspace-restart-handoff.test.js tests/flow-wogi-claude-wrapper.test.js tests/flow-wave1-integrations.test.js tests/flow-wave2-integrations.test.js && NODE_ENV=test node tests/run-quality-gates.test.js",
     "test:syntax": "find scripts/ lib/ -name '*.js' -not -path '*/node_modules/*' -exec node --check {} +",
     "lint": "eslint scripts/ lib/ tests/",
     "lint:ci": "eslint scripts/ lib/ tests/ --max-warnings 0",

package/scripts/flow-extraction-review.js

@@ -390,6 +390,46 @@ function getConfirmedTasks() {
     }));
 }
 
+/**
+ * v2.24.0 — Export confirmed items as an "Item Manifest" compatible with
+ * /wogi-story's item-reconciliation gate (wf-63c0f4cc). Downstream callers
+ * (/wogi-story, /wogi-epics, /wogi-feature, /wogi-plan) can pass this to
+ * their P0 gates as `fullInput` and mark `bypassLongInput: true` so the
+ * re-extraction loop is skipped.
+ *
+ * @returns {{items: Array<string>, count: number, bypassLongInput: true, sourceSessionId: string, intentBootstrapScheduled: boolean}}
+ */
+function exportAsItemManifest() {
+  const session = loadReviewSession();
+  if (!session) throw new Error('No review session active');
+  if (!session.completeness_confirmed) {
+    throw new Error('Cannot export manifest: review not yet confirmed as complete');
+  }
+
+  const items = session.items
+    .filter(i => i.review_status === 'confirmed')
+    .map(i => (i.text || '').trim())
+    .filter(Boolean);
+
+  // Coordinate with Intent Bootstrap (see flow-story-gates.coordinateIntentBootstrap)
+  // so /wogi-start doesn't re-prompt if the user already scheduled bootstrap via
+  // /wogi-story during this session.
+  let intentBootstrapScheduled = false;
+  try {
+    const gates = require('./flow-story-gates');
+    const result = gates.coordinateIntentBootstrap();
+    intentBootstrapScheduled = !!(result && result.scheduled);
+  } catch (_err) { /* non-critical */ }
+
+  return {
+    items,
+    count: items.length,
+    bypassLongInput: true,
+    sourceSessionId: session.id || null,
+    intentBootstrapScheduled
+  };
+}
+
 // =============================================================================
 // DISPLAY HELPERS
 // =============================================================================
@@ -779,6 +819,7 @@ module.exports = {
 
   // Get results
   getConfirmedTasks,
+  exportAsItemManifest,  // v2.24.0 — Wave 2 alignment with /wogi-story P0 gates
 
   // Display
   formatReviewStatus,
@@ -858,6 +899,16 @@ if (require.main === module) {
       }
       break;
 
+    case 'manifest':
+      // v2.24.0 — export Item Manifest for downstream /wogi-story coordination
+      try {
+        const manifest = exportAsItemManifest();
+        console.log(JSON.stringify(manifest, null, 2));
+      } catch (err) {
+        console.error(`${c.red}✗ ${err.message}${c.reset}`);
+      }
+      break;
+
     default:
       console.log('Extraction Review Module');
       console.log('Commands:');

package/scripts/flow-morning.js

@@ -377,9 +377,31 @@ function collectBriefingData() {
     git: gitStatus,
     progress: progressData,
     metrics: sessionState.metrics,
-    suggestedPrompt: null // Will be generated if enabled
+    suggestedPrompt: null, // Will be generated if enabled
+    workspaceOverdue: null,   // v2.23.0 — populated below when manager mode
+    honestyHits: []           // v2.23.0 — populated below
   };
 
+  // v2.23.0 — Workspace dispatch surfacing (manager mode only).
+  // If the user is working inside a workspace manager session, surface any
+  // overdue or restart-gap-lost dispatches so the morning briefing catches
+  // what the last manager turn would have caught. Fail-open.
+  try {
+    if (process.env.WOGI_WORKSPACE_ROOT) {
+      const { buildOverdueContext } = require('./hooks/core/overdue-dispatches');
+      const ctx = buildOverdueContext();
+      if (ctx) briefing.workspaceOverdue = ctx;
+    }
+  } catch (_err) { /* non-critical */ }
+
+  // v2.23.0 — Completion-claim honesty scan.
+  // Catches done-word-in-notes-while-status-partial and similar
+  // contradictions across ready.json (uses the honesty-infra from 2026-04-16).
+  try {
+    const { checkCompletionClaimHonesty } = require('./flow-health');
+    briefing.honestyHits = checkCompletionClaimHonesty();
+  } catch (_err) { /* non-critical */ }
+
   // Generate suggested prompt if enabled
   if (morningConfig.generatePrompt !== false) {
     briefing.suggestedPrompt = generateSuggestedPrompt(briefing);
@@ -532,6 +554,30 @@ function printBriefing(briefing) {
     }
   }
 
+  // v2.23.0 — Workspace dispatch issues (manager mode)
+  if (briefing.workspaceOverdue) {
+    printSection('WORKSPACE DISPATCH ISSUES');
+    for (const line of briefing.workspaceOverdue.split('\n')) {
+      console.log(`  ${line}`);
+    }
+    console.log('');
+  }
+
+  // v2.23.0 — Completion-claim honesty contradictions
+  if (Array.isArray(briefing.honestyHits) && briefing.honestyHits.length > 0) {
+    printSection('HONESTY CHECK — claim/state contradictions');
+    console.log(`  ${color('yellow', '\u26a0')} ${briefing.honestyHits.length} task${briefing.honestyHits.length === 1 ? '' : 's'} with claim-vs-state mismatch`);
+    for (const h of briefing.honestyHits.slice(0, 5)) {
+      const snippet = String(h.snippet || '').slice(0, 80);
+      console.log(`    ${color('dim', `${h.id} (class ${h.class}, field ${h.field}): "${snippet}..."`)}`);
+    }
+    if (briefing.honestyHits.length > 5) {
+      console.log(`    ${color('dim', `... and ${briefing.honestyHits.length - 5} more`)}`);
+    }
+    console.log(`  ${color('dim', 'Review: flow health')}`);
+    console.log('');
+  }
+
   // Changes since last session
   if (morningConfig.showChanges !== false) {
     const changes = briefing.changesSinceLastSession;

package/scripts/flow-session-end.js

@@ -580,6 +580,53 @@ function runCompletionContradictionScan() {
   }
 }
 
+/**
+ * v2.23.0 — Workspace session-end handoff.
+ *
+ * When running inside a workspace manager session, write a structured
+ * `heads-up` message to .workspace/messages/ so workers (and subsequent
+ * manager sessions) can see that the manager is closing cleanly. Pairs
+ * with the worker-ready / worker-stopped / lost-dispatches protocol
+ * from 2.22.x — gives workers a definitive "manager went away, no new
+ * dispatches coming" signal instead of silent absence.
+ *
+ * Fail-open: any error is logged in DEBUG mode but never blocks session-end.
+ */
+function writeWorkspaceSessionEndMessage() {
+  const workspaceRoot = process.env.WOGI_WORKSPACE_ROOT;
+  if (!workspaceRoot) return;
+  const repo = process.env.WOGI_REPO_NAME;
+  // Only manager-mode sessions emit this signal. Workers use their own
+  // Stop-hook worker-stopped message (see lib/workspace-messages.js).
+  if (repo && repo !== 'manager') return;
+
+  try {
+    const messagesLib = path.resolve(__dirname, '..', 'lib', 'workspace-messages.js');
+    const { createMessage, saveMessage } = require(messagesLib);
+    const msg = createMessage({
+      from: repo || 'manager',
+      to: 'all',
+      type: 'heads-up',
+      subject: 'Manager session ended',
+      body: [
+        'The workspace manager session has ended cleanly.',
+        'No new dispatches will arrive until a new manager session starts.',
+        'Workers finishing current tasks can safely end-of-turn and restart.'
+      ].join('\n'),
+      priority: 'medium',
+      actionRequired: false
+    });
+    saveMessage(workspaceRoot, msg);
+    if (process.env.DEBUG) {
+      console.error(`[session-end] Wrote manager-session-ended message: ${msg.id}`);
+    }
+  } catch (err) {
+    if (process.env.DEBUG) {
+      console.error(`[session-end] Workspace session-end message failed (non-fatal): ${err.message}`);
+    }
+  }
+}
+
 /**
  * v1.7.0: Archive request log if threshold exceeded
  */
@@ -1256,6 +1303,9 @@ async function main() {
   // hotfixes[] evidence. Surface-and-prompt, not hard-fail (calibration first).
   runCompletionContradictionScan();
 
+  // v2.23.0 — Workspace session-end message (manager mode only).
+  writeWorkspaceSessionEndMessage();
+
   console.log('');
 
   // Offer to push