fix(bmm): address review batch on bmad code-review/quick-dev skills

- step-02-review.md: static prefilter now runs BOTH the policy.toml
  [verify] gates AND the spec's ## Verification commands (deduped),
  instead of "first match wins". The orchestrator runs only policy.toml
  itself, so spec-specific checks were otherwise skipped whenever
  policy.toml existed (augment).
- step-03-triage.md: guard the "Clean review — all layers passed"
  announcement on {failed_layers} being empty, and fix the stale "Step 3
  already warned" reference (the warning lives in step 6) — resolves the
  contradiction with step 6 (augment).
- deferred-work-format.md (quick-dev): add the sibling-copy sync note to
  mirror the bmad-code-review copy (coderabbit).
- automation-mode.md: catalogue the spec-write-failure and
  verification-failure escalation types in the decision table so every
  emitted type is documented (coderabbit).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

src/bmm-skills/4-implementation/bmad-code-review/steps/step-02-review.md

@@ -16,10 +16,14 @@ failed_layers: '' # set at runtime: comma-separated list of layers that failed o
 
 0. **Static prefilter** (`{auto_mode}` only — skip entirely in interactive runs). Before any LLM review, run the project's deterministic
    checks — they are free, precise findings the hunters should not have to
-   rediscover. Resolve the command list in this order, first match wins:
+   rediscover. Build the command list from BOTH sources below (the orchestrator
-   1. `[verify] commands` in `{project-root}/.automator/policy.toml` (if the file exists)
+   runs only the policy.toml gates itself, never the spec's, so the spec's
-   2. the `## Verification` commands in `{spec_file}` (if `{review_mode}` = `"full"`)
+   story-specific checks must run here or nowhere), then run each unique command
-   3. none found — skip this instruction silently.
+   once — dedupe identical command strings:
+   - `[verify] commands` in `{project-root}/.automator/policy.toml` (if the file exists) — the orchestrator's project-wide gates
+   - the `## Verification` commands in `{spec_file}` (if `{review_mode}` = `"full"`) — story-specific checks
+
+   If neither source yields a command, skip this instruction silently.
 
    Record each failing command as a finding with `source: static` (title = the
    command, detail = the failure output tail). Summarize failures in one line

src/bmm-skills/4-implementation/bmad-code-review/steps/step-03-triage.md

@@ -62,7 +62,7 @@
 
 6. If `{failed_layers}` is non-empty, report which layers failed before announcing results. If zero findings remain after dropping dismissed AND `{failed_layers}` is non-empty, warn the user that the review may be incomplete rather than announcing a clean review.
 
-7. If zero findings remain after triage (all rejected or none raised): state "✅ Clean review — all layers passed." (Step 3 already warned if any review layers failed via `{failed_layers}`.)
+7. If zero findings remain after triage (all rejected or none raised) AND `{failed_layers}` is empty: state "✅ Clean review — all layers passed." (If any layers failed, step 6 already issued the incompleteness warning instead of a clean announcement.)
 
 
 ## NEXT

src/bmm-skills/4-implementation/bmad-quick-dev/automation-mode.md

@@ -109,6 +109,8 @@ overflow choose `[S] Split` per the decision table.
 | step-01/02 unclear intent after investigation          | Escalate `CRITICAL` (`type: intent-gap`). Do not fantasize requirements.                                                                                                             |
 | step-02 token budget exceeded                          | Choose **[S] Split** (defer secondary scope per `./deferred-work-format.md`).                                                                                                        |
 | step-02 CHECKPOINT 1                                   | Perform the self-review against the READY FOR DEVELOPMENT standard, fix what it surfaces, then auto-approve: set status `ready-for-dev`, lock the frozen block, continue to step-03. |
+| step-02 CHECKPOINT 1 spec preflight fails              | If `{spec_file}` is missing or empty after the write, escalate `CRITICAL` (`type: spec-write-failure`) instead of approving.                                                         |
+| step-auto-finalize verification cannot pass            | If a spec `## Verification` command cannot pass without violating the frozen intent, escalate `CRITICAL` (`type: verification-failure`) instead of finalizing.                       |
 | step-03 missing/empty spec precondition                | Escalate `CRITICAL` (`type: missing-spec`).                                                                                                                                          |
 | step-04 no sub-agents → "generate prompt files & HALT" | Only reachable when `$BMAD_AUTO_SKIP_REVIEW` is set. Sub-agents are pre-authorized — run the three reviewers inline; never generate prompt files or HALT.                            |
 | step-04 `intent_gap` finding (loop back to human)      | Revert the code changes, then escalate `CRITICAL` (`type: intent-gap`). Do not infer intent.                                                                                         |

src/bmm-skills/4-implementation/bmad-quick-dev/deferred-work-format.md

@@ -2,7 +2,10 @@
 
 Canonical entry format for `{implementation_artifacts}/deferred-work.md`.
 Used (in `bmad-auto` automation mode) by bmad-quick-dev (multi-goal splits,
-token splits, review defers) and bmad-code-review (defer findings). The file is
+token splits, review defers) and bmad-code-review (defer findings). This copy
+lives in the bmad-quick-dev skill; bmad-code-review ships a sibling copy of the
+same entry format (its surrounding prose differs slightly) — keep the format
+itself in sync across the two when it changes. The file is
 append-only — never rewrite or delete existing entries. (One exception: freeform
 pre-DW-format content from older projects is rewritten wholesale into canonical
 entries by a `bmad-auto sweep` migration run — an orchestrator-side process