test: remove the AI scorer that grades skill prose for clarity (freeze repair, GST-1371) (#4009)

* test: remove the skill-prose clarity LLM judges (GST-1371)

Removes the 21 eval cases where an LLM scored a skill's instruction text for
clarity, completeness and actionability, with their registrations, input
builders, response cache, calibration corpora, the cookie-workflow manual-review
approval and scripts/judge-controls.ts. The three remaining skill-llm-eval
judges keep their prompts, thresholds and samples; the file's supervision wall
shrinks to fit them. Outcome judges are untouched.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* test: pin the retired judge-control script and cookie manual-review record as deleted (GST-1371)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* test: register the manual-review fixture as free (GST-1371)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* test: keep the orphaned clarity-judge calibration corpora for a follow-up deletion (GST-1371)

Deleting the four corpus directories alone puts the diff over the /ship review
ceiling; they serve no configuration now and go in a follow-up.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* docs(todos): file the clarity-judge removal follow-ups (GST-1371)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* test: keep skill-bundle section paths POSIX so references dedupe on Windows (GST-1371)

windows-scoped-shard (run 41918796523) failed 'cross-skill references retain
complete bytes once': path.join gave the section entry a backslash path, so a
reference to the same section was not recognized as already bundled.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

---------

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

.github/cookie-workflow-manual-review.json

@@ -1,13 +0,0 @@
-{
-  "schema_version": 1,
-  "test_name": "setup-browser-cookies/SKILL.md workflow",
-  "prompt_sha256": "7f7f76f49912818d51f25c86d686aab6ad2c95ccdea60ac912c1e4c8b6936e5f",
-  "prompt_bytes": 10159,
-  "model": "claude-fable-5-1",
-  "max_tokens": 8192,
-  "thresholds": { "clarity": 4, "completeness": 3, "actionability": 4 },
-  "approved_by": "garrytan",
-  "approved_at": "2026-09-24",
-  "approval_url": "https://github.com/garrytan/gstack/pull/2964#issuecomment-5822514476",
-  "reason": "Maintainer-approved manual review of this exact cookie workflow after empty provider refusals. No automated quality score or pass credit; other errors and changed inputs remain blocking."
-}

.github/cookie-workflow-manual-review.md

@@ -1,22 +0,0 @@
-# Cookie workflow manual-review exception
-
-`cookie-workflow-manual-review.json` records the maintainer's approval for one
-exact cookie-workflow judge request. It is not generated content. Do not update
-its hash, model, budget, thresholds, or provenance just to make a changed test
-pass; a changed request needs a new explicit review and approval.
-
-The ordinary judge request still runs. Only an explicit provider refusal with
-complete request/response identifiers, zero output tokens, and no text blocks
-can use this approval. Low scores, malformed output or evidence, other errors,
-timeouts, and late or superseded attempts remain failures. Every other case
-remains subject to its existing gate.
-
-Manual acceptance is first-attempt-only: a retry refusal cannot erase an earlier
-scored failure, timeout, or other error. Normal configured retries are unchanged.
-
-The collector preserves `passed: false`, `execution: executed`, the refusal,
-and the manual approval, without a score or score-cache receipt. Reports count
-manual acceptance separately from automated passes and failures; “executed”
-counts the actual provider request, not a completed scored evaluation. Historical
-records retain that distinction. CI also checks manual claims against the
-current source and committed approval before accepting them.