wicked-garden 12.36.0 → 12.37.0

package/.claude-plugin/marketplace.json

@@ -1 +1 @@
-{"name": "wicked-garden", "owner": {"name": "Mike Parcewski"}, "plugins": [{"name": "wicked-garden", "source": "./", "description": "Your coding agent already plans and swarms \u2014 Wicked Garden is the curated toolkit for what it can't: proof-not-claims gates, code links grep can't see, deterministic refactor, cross-session memory, real multi-model second opinions. Don't fight the harness; fill its gaps. The gate needs wicked-vault (a self-contained infra peer, installed directly) and re-derives through garden's built-in engine; estate/bus are opt-in layers (wicked-testing and wicked-brain are retired \u2014 QE ships in-catalog as the qe domain, memory lives in wicked-estate).", "version": "12.36.0"}], "version": "12.36.0"}
+{"name": "wicked-garden", "owner": {"name": "Mike Parcewski"}, "plugins": [{"name": "wicked-garden", "source": "./", "description": "Your coding agent already plans and swarms \u2014 Wicked Garden is the curated toolkit for what it can't: proof-not-claims gates, code links grep can't see, deterministic refactor, cross-session memory, real multi-model second opinions. Don't fight the harness; fill its gaps. The gate needs wicked-vault (a self-contained infra peer, installed directly) and re-derives through garden's built-in engine; estate/bus are opt-in layers (wicked-testing and wicked-brain are retired \u2014 QE ships in-catalog as the qe domain, memory lives in wicked-estate).", "version": "12.37.0"}], "version": "12.37.0"}

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "wicked-garden",
-  "version": "12.36.0",
+  "version": "12.37.0",
   "description": "Your coding agent's harness already plans and swarms; Wicked Garden is the curated toolkit for what it can't do alone. Proof, not claims \u2014 gates re-derive 'done' through the internal loom engine (fail-closed, independently attested via wicked-vault), so a self-asserted pass can't lie its way green. Relationships grep can't see \u2014 the wicked-estate code graph + injected bus/dispatch/capability/archetype edges power blast-radius & lineage. Plus deterministic multi-file refactor (wicked-patch), cross-session memory + knowledge (the mem domain over wicked-estate), real multi-model second opinions (jam:council), portable expertise as on-demand skill-refs, and evidence-gated testing (the in-catalog qe domain). It reads the shape of your work to apply the right rigor \u2014 steering, not a pipeline \u2014 then gets out of the way. The compiler stamps a self-contained evidence gate into any repo that runs with no wicked-garden installed. Don't fight the harness; fill its gaps. The evidence gate requires wicked-vault, a self-contained infra peer wicked-garden installs directly; wicked-bus is an opt-in layer \u2014 install what you'll use, the toolkit works without the rest.",
   "wicked_vault_version": "^0.5.0",
   "wicked_ledger_version": "^0.4.0",

package/CHANGELOG.md

@@ -2,6 +2,11 @@
 
 ## [Unreleased]
 
+## [12.37.0] — 2026-09-14
+
+### Changed
+- **Evaluator verdicts: one grammar — `VERDICT: PASS` or `VERDICT: FAIL`, one plain-text line, the LAST line of the message (FIX-IT-ALL L6-0; the garden half of D-9, mirrored by wicked-crew's qe review text and wicked-core's evaluator-verdict carrier).** The review surfaces said APPROVE / CONDITIONAL / REJECT in one place (`wicked-garden-governed-worker`), `VERDICT={PASS|CONDITIONAL|FAIL} MODE=produced-test` in another (the `qe` router's `review` action), `{PASS|CONDITIONAL|FAIL|SKIP}` in a third (`qe-security-test-engineer`), TWO verdict lines in one message (`qe-test-code-quality-auditor` §8 + §9) and no verdict line at all in an evaluator skill (`domain-coverage`) — a gate cannot read five spellings. wicked-core's evaluator gate (wave 3, core-ts 0.7.27) reads the LAST `^VERDICT[:=]` line of an evaluator unit's output and passes ONLY on the token `PASS`; any other token — or no line — parks the run at the human gate, never a false pass. Every garden text surface now asks for exactly that: `governed-worker` (evaluator method; output contract item 6 "the contract line — the LAST line of your message"; a self-reviewing seat writes no line and the gate treats that as FAIL), `qe/refs/review.md` (vocabulary + the produced-test dispatch — `MODE: produced-test` on its own line), `qe-test-code-quality-auditor` (the §8 audit line becomes `AUDIT: {p0}/{p1}/{p2} REVIEWER: … RUN_ID: …` with no VERDICT token — a smell audit certifies nothing; §9 carries the ONE contract line with `MODE:` / `REVIEWER:` / `RUN_ID:` on their own lines), `qe-security-test-engineer` (`SKIP` → `VERDICT: FAIL` + `reason: not applicable: <why>` — a review with no security surface parks at the gate instead of passing silently), `qe-a11y-test-engineer` / `qe-ai-feature-test-engineer` / `qe-snapshot-hygiene-auditor` (the footer is the grammar with `REVIEWER:` / `RUN_ID:` on their own lines; CONDITIONAL stays a ledger RECORD value, the output line is FAIL with the conditions listed above it), `domain-coverage` (gains the line: PASS iff `coverage == 1.0`, zero unaccounted and no `BLOCK`), `qe/refs/accept.md` (the acceptance reviewer writes PASS|FAIL; `VERDICT_TO_STATUS` is unchanged — PARTIAL / CONDITIONAL / INCONCLUSIVE are legacy record values readers keep mapping) and `wickedizer/refs/patterns.md` (the comparison template's `VERDICT:` heading is prose again). **Behaviour change (D-9, text half):** CONDITIONAL and SKIP are no longer verdicts — unmet conditions and not-applicable reviews FAIL into the human gate; the wicked-ledger `Verdict` enum and the DomainStore `verdict:` records are untouched (record vocabulary). **Lint:** `tests/test_skill_portability.py` gains the garden-only token `verdict-spelling` — every line-leading `VERDICT[:=]` line under `skills/**/*.md` (fenced blocks included) must be `VERDICT: PASS`, `VERDICT: FAIL` or the template `VERDICT: PASS|FAIL`; nothing that is not line-leading is judged. The legacy `VERDICT=… REVIEWER=… RUN_ID=…` footers of seven `qe-*` specialists already parse as PASS/FAIL and are tolerated ONLY in the files listed under `verdict_spelling.legacy_footer_files` in `tests/portability_rules.garden.json`; a listed file with no footer left fails the build, so the qe batches that rewrite them shrink the list in the same change. **Mixed generations:** a skills generation published from garden ≤ 12.36.0 still writes the old words and gates every evaluator unit under core-ts 0.7.27 — publish garden ≥ 12.37.0 first.
+
 ## [12.36.0] — 2026-09-13
 
 ### Fixed
@@ -852,3 +857,6 @@ through `v9.2.18`. Major themes across that span:
   condition-resolution mechanism, scope-delta HITL trigger.
 
 For each of those releases, see the corresponding tag in git history.
+
+[Unreleased]: https://github.com/mikeparcewski/wicked-garden/compare/v12.37.0...HEAD
+[12.37.0]: https://github.com/mikeparcewski/wicked-garden/compare/v12.36.0...v12.37.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wicked-garden",
-  "version": "12.36.0",
+  "version": "12.37.0",
   "type": "module",
   "description": "Curated toolkit for what AI coding agents can't do alone: evidence-gated work, codegraph relationships, deterministic multi-file refactor, multi-model second opinions, and cross-session memory. Installs as a Claude Code plugin.",
   "keywords": [

package/skills/domain-coverage/SKILL.md

@@ -77,3 +77,8 @@ Then one threat per line: `[RISK|BLOCK|NOTE] <location> — <finding>`.
 
 A `BLOCK` means the domain model must not proceed to build until resolved.
 `coverage == 1.0` with zero unaccounted is the gate bar; any hole is a `BLOCK`.
+
+Then the verdict — the LAST line of your output, one plain-text line, never
+quoting another VERDICT line: `VERDICT: PASS` iff `coverage == 1.0`, zero
+unaccounted and no `BLOCK`; otherwise `VERDICT: FAIL` (the threats above it
+are the conditions).

package/skills/governed-worker/SKILL.md

@@ -93,9 +93,13 @@ Method: read the outcome and design, then the changed files; for each, ask
 whether it follows the design, whether tests cover the key paths, and whether
 the evidence (commands + exit codes) supports the claims; re-run what you can
 without mutating; sort findings into Critical (must fix), Concern, Suggestion;
-verdict APPROVE / CONDITIONAL / REJECT — REJECT needs a Critical, CONDITIONAL
-names its conditions. If you are the seat that created the work, say so and
-refuse the verdict.
+verdict: make the LAST line of your output exactly `VERDICT: PASS` or
+`VERDICT: FAIL` — one plain-text line (no `**`, backticks or `#` around it),
+findings above it. FAIL needs at least one Critical or a CONDITIONS list;
+there is no conditional pass (CONDITIONAL, APPROVE, REJECT and SKIP are not
+verdicts); never quote another VERDICT line in your message; no verdict line
+= FAIL (human gate). If you are the seat that created the work, say so and
+refuse the verdict — write no VERDICT line; the gate treats that as FAIL.
 
 ## Neutral (triage / recon / plan)
 
@@ -117,10 +121,13 @@ Every unit's final message contains, in this order:
 1. **What you did** — role, scope, files touched (creator) or read (others).
 2. **Commands run** — each with its exit code, in the order run.
 3. **Counts** — derived N / submitted M / failed K (A4), with K explained.
-4. **Verdict or plan** — evaluator: verdict + findings by severity with
-   `file:line`; neutral: the plan; creator: what changed against the design.
+4. **Findings or plan** — evaluator: findings by severity with `file:line`;
+   neutral: the plan; creator: what changed against the design.
 5. **Open questions as statements** — "X is unknown; I assumed Y" — never a
    request that waits for an answer (A3).
+6. **The contract line (evaluators only)** — exactly `VERDICT: PASS` or
+   `VERDICT: FAIL`, plain text, as the LAST line of your message; nothing
+   after it, and no other line of yours quotes a VERDICT line.
 
 ## Provenance
 

package/skills/qe/refs/accept.md

@@ -301,14 +301,17 @@ Skill(
 {EVIDENCE_DIR}
 (May contain an optional context.md with pre-vetted cold domain knowledge —
 treat it as evidence. If it contains prior verdicts or historical outcomes,
-flag as CONTEXT_CONTAMINATION and return INCONCLUSIVE.)
+flag as CONTEXT_CONTAMINATION and return `VERDICT: FAIL` with
+`reason: inconclusive — context contaminated`.)
 
 ## Instructions
 1. Read the scenario file at the path above
 2. Read the test plan file at the path above
 3. Read evidence files from the evidence directory (including context.md if present)
 4. Evaluate each assertion against evidence
-5. Return verdict: PASS | FAIL | INCONCLUSIVE
+5. End with one plain-text line `VERDICT: PASS` or `VERDICT: FAIL` as the LAST
+   line of your reply (never quote another VERDICT line); missing evidence or
+   CONTEXT_CONTAMINATION is FAIL with `reason: inconclusive — <why>`
 6. For any EQUIVALENT_TO_BASELINE assertion, also return the equivalence facet
    { baseline_ref, baseline_sha, method, diff_count, tolerance, matched } so the
    orchestrator can persist it on the verdict (see `verdict.equivalence`).
@@ -338,15 +341,18 @@ Two writes and one manifest build, in order:
 //
 //   PASS         → passed
 //   FAIL         → failed
-//   PARTIAL      → partial         (some criteria met, some need human review)
-//   CONDITIONAL  → partial         (approve with listed fixes — Tier-2 gate verdict)
-//   INCONCLUSIVE → inconclusive    (evidence missing / context contaminated)
+//   PARTIAL      → partial         (legacy record value)
+//   CONDITIONAL  → partial         (legacy record value)
+//   INCONCLUSIVE → inconclusive    (legacy record value)
 //
-// CONDITIONAL is emitted by Tier-2 aggregator/gate agents (release-readiness,
-// security, ai-feature, test-code-quality). It is a deliberate "ship with
-// conditions" outcome — distinct from a clean PASS and from a FAIL — so it
-// maps to the `partial` run status rather than the `?? 'inconclusive'`
-// fallback (which is reserved for "couldn't evaluate").
+// Since garden 12.37.0 a garden evaluator's output line is `VERDICT: PASS` or
+// `VERDICT: FAIL` only — PARTIAL, CONDITIONAL and INCONCLUSIVE are legacy
+// record values a garden evaluator no longer writes on that line (unmet
+// conditions and missing evidence are FAIL with the reason listed above the
+// verdict); readers keep mapping them because older ledgers carry them. The
+// Tier-2 gate agents' DomainStore records may still carry CONDITIONAL as a
+// record value — the `partial` run status, distinct from the `?? 'inconclusive'`
+// fallback (reserved for "couldn't evaluate").
 //
 // `errored` and `skipped` remain reachable only via the run lifecycle
 // (executor crash / stale-run sweep, all-steps-skipped) — never from a verdict.
@@ -383,7 +389,7 @@ const reviewerEquivalenceFacet = reviewerResponse.equivalence ?? null;
 const reviewerEquivalence = reviewerEquivalenceFacet ?? null;
 const verdictRecord = store.create('verdicts', {
   run_id: run.id,
-  verdict: reviewerVerdict,            // 'PASS' | 'FAIL' | 'PARTIAL' | 'CONDITIONAL' | 'INCONCLUSIVE'
+  verdict: reviewerVerdict,            // 'PASS' | 'FAIL' from the reviewer's VERDICT line (PARTIAL / CONDITIONAL / INCONCLUSIVE: legacy record values readers still map)
   evidence_path: EVIDENCE_DIR,
   reviewer: 'acceptance-test-reviewer',
   reason: reviewerSummary,
@@ -426,7 +432,7 @@ emitBusEvent('wicked.test.evidence.captured', {
 ```markdown
 ## Acceptance Test Results: {scenario name}
 
-### Verdict: {PASS | FAIL | INCONCLUSIVE}
+### Verdict: {PASS | FAIL}
 
 ### Acceptance Criteria
 | Criterion | Verdict | Evidence |

package/skills/qe/refs/review.md

@@ -142,13 +142,16 @@ where the harness is runnable it re-runs the produced files itself.
    merges the PR and never pushes — the run's deliver phase (or the human)
    does; a reviewer that ships is a creator grading itself.
 
-Verdict vocabulary (`MODE=produced-test`): **`PASS`** only when 1–7 hold with
+Verdict vocabulary (`MODE: produced-test`): **`PASS`** only when 1–7 hold with
 zero findings — every `covered` claim verified at its `path:line`, every
 produced test executed and green (your re-run reproduces the record), no
-`unverified` or `failing` row; **`CONDITIONAL`** with the fixes listed when only
-P1/P2 rows fail (implementation-shaped test, padded total, silent scope gap,
-style); **`FAIL`** on any `[unexecuted-test]`, red produced test, tautological
-test, unsupported-`covered` claim, or `[scenario-defect]` oracle. Cite
+`unverified` or `failing` row; **`FAIL`** otherwise — on any
+`[unexecuted-test]`, red produced test, tautological test,
+unsupported-`covered` claim or `[scenario-defect]` oracle, AND when only P1/P2
+rows fail (implementation-shaped test, padded total, silent scope gap, style):
+list those fixes as CONDITIONS above the verdict line; there is no conditional
+pass. The verdict is one plain-text line, `VERDICT: PASS` or `VERDICT: FAIL`,
+the LAST line of the message; never quote another VERDICT line. Cite
 `file:line` for every finding. The auditor's §9 carries the same rule.
 
 Dispatch fallback as in § Dispatch block above: without a Skill tool, open the
@@ -175,7 +178,7 @@ MODE: produced-test — grade per the auditor's §9 produced-test verdict
    `unverified`; recount new vs pre-existing tests.
 2. Confirm every produced file has an execution record (file · command ·
    result). None — or a `needs-fixture` e2e never run against its fixture —
-   is VERDICT=FAIL `[unexecuted-test]`.
+   is `VERDICT: FAIL` `[unexecuted-test]`.
 3. Re-run the produced files when the harness is available; compare with
    the record.
 4. Mutate-or-reason about at least two tested behaviours; name the failing
@@ -188,10 +191,13 @@ MODE: produced-test — grade per the auditor's §9 produced-test verdict
    a mock, a snapshot of a fixture, or one that would pass with the
    implementation deleted.
 
-Return `VERDICT={PASS|CONDITIONAL|FAIL} MODE=produced-test` — PASS only with
-every claim verified at path:line, every produced test executed and green, no
-`unverified` — and per-row findings with file:line. Do NOT push, open, edit or
-merge a PR — the run's deliver phase delivers."""
+Report `MODE: produced-test` on its own line and per-row findings with
+file:line, then end with one plain-text line `VERDICT: PASS` or `VERDICT: FAIL`
+as the LAST line of your message — PASS only with every claim verified at
+path:line, every produced test executed and green, no `unverified`; anything
+less is FAIL with the conditions listed above it; never quote another VERDICT
+line. Do NOT push, open, edit or merge a PR — the run's deliver phase
+delivers."""
 )
 ```
 

package/skills/qe-a11y-test-engineer/SKILL.md

@@ -39,8 +39,11 @@ This skill is designed to run as an isolated worker; when your harness cannot fo
 You test that the UI works for people who don't use a mouse or don't see
 the screen. Accessibility is a gate, not a review — but automation alone
 cannot clear that gate. Axe-core catches roughly 30% of WCAG failures;
-the rest require a human. Your default verdict is therefore **CONDITIONAL**
-with an explicit list of unchecked manual items.
+the rest require a human. Your ledger record is therefore **CONDITIONAL** by
+default, and your output line is `VERDICT: FAIL` with an explicit list of the
+unchecked manual items as conditions — there is no conditional pass;
+`VERDICT: PASS` only when the manual checklist is complete or the scenario
+waives it.
 
 ## Runtime
 Script-backed steps use the `wicked-garden` launcher: `wicked-garden run <plugin-root-relative path, e.g. scripts/…> [args]`. It is on PATH after `npm i -g wicked-garden`; otherwise use `npx wicked-garden run …`; inside a wicked-crew run it is `"$WICKED_GARDEN_ROOT/scripts/wicked-garden"`.
@@ -225,10 +228,11 @@ refuse a verdict — a11y is a gate.
 
 - **WCAG 2.1 AA is the floor, not the ceiling.** If the scenario's
   context.md requires 2.2 AA, use the `wcag22aa` tag as well.
-- **Zero axe violations ≠ compliant.** Always render the verdict as
-  `CONDITIONAL` (approve with the manual checklist as the listed fixes)
-  unless the scenario explicitly waives manual review — never `N-A`:
-  the a11y item always applies (see the DomainStore write above).
+- **Zero axe violations ≠ compliant.** Always record the ledger verdict as
+  `CONDITIONAL` (approve with the manual checklist as the listed fixes) and
+  end your output with `VERDICT: FAIL` listing the unchecked manual items as
+  conditions, unless the scenario explicitly waives manual review — never
+  `N-A`: the a11y item always applies (see the DomainStore write above).
 - **Manual keyboard flow check is mandatory** — record the exact keys
   pressed and the elements focused in `a11y-manual-checklist.md`. If
   Playwright isn't available, leave the checklist items as unchecked
@@ -258,7 +262,8 @@ every test you produce, whether dispatched or invoked directly:
 ## 7. Output format
 
 Print a compact summary to stdout; the full detail lives in the evidence
-files. The final line MUST be a machine-readable verdict tag.
+files. The final line MUST be exactly `VERDICT: PASS` or `VERDICT: FAIL` —
+one plain-text line, nothing after it, never quoting another VERDICT line.
 
 ```
 ## A11y: {scenario.name}
@@ -266,9 +271,12 @@ target: {TARGET_URL}
 axe: {axeViolationCount} violations ({critical} critical, {serious} serious, {moderate} moderate, {minor} minor)
 pa11y: {pa11yErrorCount} errors, {pa11yWarningCount} warnings
 keyboard: {keyboardPassCount}/{keyboardTotalCount} steps reached target without a trap
-verdict: CONDITIONAL (axe+pa11y clean, manual review required)
+record: CONDITIONAL (axe+pa11y clean, manual review required)
+conditions: {unchecked manual checklist items}
 
-VERDICT=CONDITIONAL REVIEWER=wicked-garden-qe-a11y-test-engineer RUN_ID={RUN_ID}
+REVIEWER: wicked-garden-qe-a11y-test-engineer
+RUN_ID: {RUN_ID}
+VERDICT: FAIL
 ```
 
 ## Helper resolution (`{WT_LIB}`)

package/skills/qe-ai-feature-test-engineer/SKILL.md

@@ -295,6 +295,11 @@ Verdict decision tree:
 - Otherwise → **PASS** with a manual-review task (red-team coverage is
   never 100%)
 
+Those are ledger record values (`verdict: decision` below). The output line
+(§11) maps them: `PASS` → `VERDICT: PASS`; `CONDITIONAL` → `VERDICT: FAIL`
+with the monitoring / citation fixes listed as conditions — there is no
+conditional pass; `FAIL` → `VERDICT: FAIL`.
+
 ```js
 store.create("verdicts", {
   run_id: RUN_ID,
@@ -364,6 +369,9 @@ store.create("tasks", {
 
 ## 11. Output
 
+The final line is exactly `VERDICT: PASS` or `VERDICT: FAIL` — one plain-text
+line, nothing after it, never quoting another VERDICT line.
+
 ```
 ## AI: {scenario.name}
 target: {TARGET}  model_pin: {id}  temp: {t}
@@ -374,9 +382,11 @@ schema:  violations={N}/{total}
 citations: bad={pct}%
 successes: direct={N}  indirect={N}  smuggle={N}  multiturn={N}  jb_families={names}
 
-verdict: {PASS|CONDITIONAL|FAIL}  reason: {short}
+record: {PASS|CONDITIONAL|FAIL}  reason: {short}
 
-VERDICT={PASS|CONDITIONAL|FAIL} REVIEWER=wicked-garden-qe-ai-feature-test-engineer RUN_ID={RUN_ID}
+REVIEWER: wicked-garden-qe-ai-feature-test-engineer
+RUN_ID: {RUN_ID}
+VERDICT: PASS|FAIL
 ```
 
 ## Helper resolution (`{WT_LIB}`)

package/skills/qe-security-test-engineer/SKILL.md

@@ -378,6 +378,13 @@ Verdict decision tree:
 - DAST refused for safety perimeter → **SKIP** with reason
   `insufficient-authorization`
 
+Those are ledger record values (`verdict: decision` below). The output line
+(§10) maps them: `PASS` → `VERDICT: PASS`; `CONDITIONAL` → `VERDICT: FAIL`
+with the Medium findings listed as conditions; `SKIP` → `VERDICT: FAIL` with
+`reason: not applicable: <why>`; `FAIL` → `VERDICT: FAIL`. There is no
+conditional pass and no skipped pass on the output line — a review with no
+security surface parks at the human gate rather than passing silently.
+
 ```js
 store.create("verdicts", {
   run_id: RUN_ID,
@@ -453,7 +460,8 @@ Do NOT touch `runs` status directly — the orchestrator owns that field.
 ## 12. Output
 
 Print a compact summary to stdout; full detail lives in the evidence
-files. Final line is a machine-readable verdict tag.
+files. The final line is exactly `VERDICT: PASS` or `VERDICT: FAIL` — one
+plain-text line, nothing after it, never quoting another VERDICT line.
 
 ```
 ## Security: {scenario.name}
@@ -465,9 +473,11 @@ authz:   idor={pass|FAIL} role-esc={pass|FAIL} jwt-none={pass|FAIL} session-fix=
 severity: critical={N} high={N} medium={N} low={N} info={N}
 asvs: exercised={N} not_exercised={N}
 
-verdict: {PASS|CONDITIONAL|FAIL|SKIP}  reason: {short}
+reason: {short — "not applicable: <why>" when nothing was in scope}
 
-VERDICT={PASS|CONDITIONAL|FAIL|SKIP} REVIEWER=wicked-garden-qe-security-test-engineer RUN_ID={RUN_ID}
+REVIEWER: wicked-garden-qe-security-test-engineer
+RUN_ID: {RUN_ID}
+VERDICT: PASS|FAIL
 ```
 
 ## Helper resolution (`{WT_LIB}`)

package/skills/qe-snapshot-hygiene-auditor/SKILL.md

@@ -233,9 +233,12 @@ findings. Return a PASS with `reason: "no snapshots found"` and
 
 ## 7. Non-negotiable rules
 
-- **Default verdict is CONDITIONAL.** This is a reviewer tool, not a
-  gate. A FAIL is reserved for severe rubber-stamp counts or catastrophic
-  dead-file counts.
+- **Default ledger verdict is CONDITIONAL; the output line is then
+  `VERDICT: FAIL` with the top-N to re-review as conditions.** This is a
+  reviewer tool, not a gate — but there is no conditional pass on the output
+  line: `VERDICT: PASS` only for a clean tree (or the `ERR_NO_SNAPSHOTS_FOUND`
+  noop). The ledger's FAIL is reserved for severe rubber-stamp counts or
+  catastrophic dead-file counts.
 - **Rubber-stamp detection requires git history.** Shallow clones with
   depth < `rubber_stamp_window + 5` trigger `ERR_GIT_LOG_FAILED`; do
   not guess.
@@ -248,6 +251,9 @@ findings. Return a PASS with `reason: "no snapshots found"` and
 
 ## 8. Output
 
+The final line is exactly `VERDICT: PASS` or `VERDICT: FAIL` — one plain-text
+line, nothing after it, never quoting another VERDICT line.
+
 ```
 ## Snapshot audit: {scenario.name}
 target_dirs: {target_dirs}
@@ -267,7 +273,9 @@ top-5 (full list in snapshot-top-n.csv):
   score=3 tests/cassettes/legacy/old.yaml     (dead)                              action=delete
   ...
 
-VERDICT={CONDITIONAL|FAIL} REVIEWER=wicked-garden-qe-snapshot-hygiene-auditor RUN_ID={RUN_ID}
+REVIEWER: wicked-garden-qe-snapshot-hygiene-auditor
+RUN_ID: {RUN_ID}
+VERDICT: PASS|FAIL
 ```
 
 ## Helper resolution (`{WT_LIB}`)

package/skills/qe-test-code-quality-auditor/SKILL.md

@@ -307,9 +307,12 @@ top-5 (full list in test-quality-top-n.csv):
   P1 tests/api/retry.test.ts:88   hardcoded-sleep "setTimeout(() => ..., 5000)"
   ...
 
-VERDICT={CONDITIONAL|FAIL} REVIEWER=wicked-garden-qe-test-code-quality-auditor RUN_ID={RUN_ID}
+AUDIT: {p0}/{p1}/{p2} REVIEWER: wicked-garden-qe-test-code-quality-auditor RUN_ID: {RUN_ID}
 ```
 
+The audit certifies nothing, so it writes no VERDICT line — the `AUDIT:` line
+is its report line; the ledger record (§5) carries the CONDITIONAL / FAIL.
+
 ## 9. Produced-test review (a PLAN + the tests it claims)
 
 When the qe `review` action dispatches you in `produced-test` mode (`MODE:
@@ -364,16 +367,23 @@ PLAN; the author's "all green" is a claim.
   mutated-or-reasoned with a nameable failing mutation each (duty 3); e2e
   oracles verified (duty 5); `not covered` rows account for the intent
   (duty 6); and the §2 detectors report P0 = P1 = P2 = 0 on the produced files.
-- **`CONDITIONAL`** — no P0, but any P1/P2 (implementation-shaped test, padded
-  total, silent scope gap, style smells): approve with the fixes listed.
+- **`FAIL` with CONDITIONS** — no P0, but any P1/P2 (implementation-shaped
+  test, padded total, silent scope gap, style smells): list the fixes as
+  conditions above the verdict line. There is no conditional pass — the ledger
+  record (§5) may still say `CONDITIONAL`; the output line says `FAIL`.
 - **`FAIL`** — any P0: `[unexecuted-test]`, a red produced test, an
   unsupported `covered` claim, a tautological test, or an impossible oracle
   `[scenario-defect]`.
 
-Report line for this mode (the §8 line stays for the audit):
+Report block for this mode — the §8 `AUDIT:` line stays for the audit; a
+message never carries two VERDICT lines and never quotes another one. The
+verdict is one plain-text line, the LAST line of your output:
 
 ```
-VERDICT={PASS|CONDITIONAL|FAIL} MODE=produced-test REVIEWER=wicked-garden-qe-test-code-quality-auditor RUN_ID={RUN_ID}
+MODE: produced-test
+REVIEWER: wicked-garden-qe-test-code-quality-auditor
+RUN_ID: {RUN_ID}
+VERDICT: PASS|FAIL
 ```
 
 ## Helper resolution (`{WT_LIB}`)

package/skills/wickedizer/refs/patterns.md

@@ -128,7 +128,7 @@ Side-by-side evaluation.
 | Time | X weeks | Y weeks |
 | Risk | Low | Medium |
 
-VERDICT: [Which wins and why]
+Verdict — [which wins and why]
 ```
 
 ### The Sequence