Build verification review skill

Use after an Auto build phase ({step}b) claims deliverables are complete — before marking the phase DONE in {{docs.roadmap_path}}. Invoke: /build-verification-review.

Mandatory in every overseer install: .cursor/rules/build-verification-required.mdc (alwaysApply). Agents must not mark DONE until this skill returns pass.

This is the honesty gate: verify the implementation actually matches the frozen spec — not merely that tests are green or the agent said "done."

Purpose

Catch agent dishonesty: false completion claims, made-up behavior, tests that do not exercise real paths, docs/marketing that overstate what shipped, and scope drift from the frozen contract.

This did not get lost — it was always the intent of SD-3 (Thinking → Auto). The kit shipped:

Tests prove code runs; this skill proves the code matches the frozen spec and the agent was honest about what landed.

When to use

• Auto phase {step}b complete; tests reported green
• Before ROADMAP row → DONE
• Before Tier-3 merge to {{vcs.git.main_branch}}
• Especially: video pipelines, billing, user-facing outputs — anywhere agents previously "approved" falsely

Model

Always thinking-high. When honesty.require_independent_second_reviewer is warn or require, this review must run in a second chat or separate verifier runtime — not the session that built. A builder-session BV pass does not unlock DONE. When the flag is off, prefer a second session but do not fail the kit gate (process honesty only).

Inputs (read all)

1. Frozen spec artifact (frozen: true from {step}a) — ground truth
2. {{docs.roadmap_path}} — phase deliverable row
3. {{docs.handover_path}} — what the build session claims landed
4. Git diff vs feature-branch base (or muse/git log for the build commits)
5. Test files — do they assert real behavior or tautologies?
6. ok status — footprint/drift if kit files touched

Verification checklist (every item needs evidence)

Evidence table (required in skill output whenever honesty module is enabled, and recommended always):

### Evidence
| type | sha256 | ref | notes |
| --- | --- | --- | --- |
| test_output | <64 hex> | <label or path> | <optional> |

Rules for a skill verdict of pass (frozen process rules):

1. V1–V7 unchanged in meaning.
2. V8 requires citing verifiable git/test state as today.
3. When honesty.enabled: true, a pass MUST be accompanied by appending (or confirming a prior append of) a verification_evidence entry with bv_verdict: pass, matching phase_id + frozen_spec + round, and a non-empty artifacts list that includes at least one test_output artifact whose sha256 digests the test output the reviewer actually used.
4. deploy_health and screenshot artifacts are required in the entry only when the build session's claims mention deploy/health or visual/UI proof; otherwise they are omitted. The skill must not invent fake deploy/screenshot evidence.
5. When honesty.enabled: false, ledger append is skipped; V8 still requires claims↔git/test honesty in the review text (baseline without L2).
6. findings / blocked rounds MAY append verification_evidence with the corresponding bv_verdict so the chain records failed rounds; this is allowed but not required for skill progress. A later pass round is a separate append (new round).
7. V9 is DONE-unlock only — not a V1–V8 implementation check. V9 does not block a V1–V8 implementation pass and does not require the ISR line before the second session appends it. Order: V1–V8 → ISR append → Mode B when it applies → DONE. V9 is N/A (not a fail) when the flag is off or honesty is disabled. A builder-session V1–V8 pass still does not unlock DONE when ISR is warn or require.

Verdicts

Use the same escalation categories as freeze review: security, irreversible, real_money, gates_tier3.

Loop (bounded)

Same pattern as freeze-review-loop:

• Max rounds: 5 (build fixes should be smaller than spec fixes)
• Reviewer → findings → fixer (cited lines only) → re-review until pass or stop
• Do not mark DONE in ROADMAP until pass

Output format

## Build verification — <phase-id> round <N>

**Verdict:** pass | findings | blocked
**Frozen spec:** <path>
**Diff scope:** <commits or file list>

### Findings
| ID | Sev | path:line | Claim vs reality |
| --- | --- | --- | --- |
| BV1 | MAJOR | src/foo.py:42 | Spec §3.2 requires fail-closed; returns success on error |

### Honest summary (only if pass)
One paragraph: what actually shipped, backed by paths in the diff.

CLI helpers (non-substitute)

./cli/ok status
# run the phase's seven-tier test command
./cli/ok governance-sync --dry-run

CLI cannot replace thinking verification — run this skill after tests pass.

Handover integration

When {step}b completes, NEXT block should say:

Model: thinking-high.
Run build-verification-review against <frozen-spec-path> before marking <phase> DONE.

Hard stops

• No merge to {{vcs.git.main_branch}} on findings or blocked
• No waiving V8 ("tests pass therefore done") without addressing citations
• Uncited findings invalid — re-review