refactor(skills): retrofit the corpus to the ADR-0020 context contract (#129)

Retrofits all 39 skills to ADR-0020's description/body context contract, then fixes what six rounds of independent review found in that retrofit — including four ways the hot gate itself failed open.

Closes #99, #107, #108, #110, #111, #114, #115, #120.

## The retrofit (waves 1-5)

| | Start | Now |
|---|---|---|
| Description FAILs (>400 chars) | 26 | **0** |
| Body FAILs (>900 words, body-only) | 9 | **0** |
| Dangling routing targets | 2 | **0** |
| `Kyberforge.CompositionNote` | 10 | **0** |
| Preload tax | 21,005 chars | **~10,500** |

Under the 12,000-char success criterion. Per-wave detail is on #99.

## The review fixes

**The gate failed open four ways, three of them found after the retrofit shipped.** An unrecognised follower token made a dangling target vanish. A skill directory with no `SKILL.md` resolved as a valid target, so a commit could be green locally and red in a fresh clone — three existing fixtures were relying on that, one of which made the install-leak A/B pass vacuously. Then the free-standing `/name` sweep turned out to be gated on the sentence carrying a boundary marker, so route notation in any other sentence was invisible — not an ERROR, not a SUGGESTION, not an INFO — which left the documented "`/name` always blocks" promise false from a second direction. All four fixed and pinned.

**Two checks were silently not running.** `validate-provenance.sh` checks 7-8 were dead across nine skills. Waking them exposed a deeper problem: they assume `Research doc:` names a source index, but 30 of 121 entries point at topic content documents, so every new check-7 INFO was a false positive and check 8 was saved from a false-FAIL flood only by an *unannounced* skip. Checks 7/8 are now scoped to source indexes and every skip announces itself (#121).

**The retrofit's own anti-goal, four times.** ADR-0020 warns that a blunt gate gets satisfied by deleting content rather than relocating it. `diagnose` and `skill-audit` relocated prose and then read it unconditionally; `prototype` and `vale-config` deleted rules outright that survived nowhere. All four addressed.

## Verification

- `bash tests/run-tests.sh --strict` — 24 suites, 0 skipped, 0 failed
- `bash tests/run-bats.sh` — 325 tests, 0 failures
- `pre-commit run --all-files` — 17/17
- `pre-commit run --hook-stage pre-push --all-files` — 16/16, with `apm marketplace check` and `apm pack --check-clean` run against the remote, not skipped
- `scripts/skill-size-check.sh` over all 39 skills — rc 0, 0 ERROR/FAIL, SUGGESTION-only
- Preload tax measured at **10,498 chars**, max description 390 — both inside budget
- Every new test proven non-vacuous by a deliberate mutation of the behaviour it covers

**Per-commit sync, stated accurately:** the ten commits from the latest review round each pass `check-plugin-content-sync` in isolation, verified by checking each out in a detached worktree with a clean between. The earlier gitea window (`dfacf05..bedbd1d`, nine commits) does **not** — its mirror was regenerated in one batch at `bbc7300`. An earlier revision of this description claimed the property held for every commit; it does not, and a bisect through that window lands on a red commit. **Squash-merge** to collapse it, or accept that this range is not bisectable.

## Version bump

Six plugins and the catalog take a **patch**, not a minor. The branch is **89 commits — 40 `fix` / 30 `refactor` / 12 `docs` / 5 `chore` / 2 `test` — zero `feat`, zero `!`, zero `BREAKING CHANGE`** — and adds no skill, agent, command or hook. (Two earlier revisions of this section cited a stale histogram, most recently 78 commits; the figures above are measured at HEAD.) Both rules this repo ships (`forge/references/version-bump.md`, landing in this PR, and `git-commits/references/conventional-commits-spec.md`) make that a patch, and the catalog set is unchanged at 7 entries.

Not settled by that: four published files were removed from the installed tree, three moved, and `caveman` gained `disable-model-invocation`, retiring its old triggers. Under a strict reading those are major-class and currently ship under `refactor:` with no marker. Whether the deployed skill surface is a public contract is written down nowhere — worth deciding, but it outlives this PR.

## Deliberately not in scope

#112 (cherry-pick ownership, now resolved in favour of `git-commits`), #113 (`rtk git` normalisation), #116 (research fan-out), #101 (audit-skill merge), #122 (non-spec skill-root files), #123 (no PRD producer) stay open. #117 is the one worth reading: the contract's remedy is to move prose into `references/`, which is exactly where neither the size gate nor Vale looks — and the blind spot is wider than #117 currently records, since there is no root `.vale.ini` at all, so every ADR, `CONTEXT.md` and `README.md` is unlinted too.

That blind spot let this branch carry two `level: error` `Kyberforge.SentenceOpenerThereIs` violations into `references/` files it created — `provider-adapter-author/references/provider-matrix.md:31` and `agent-audit/references/finding-criteria.md:95`. Both are reworded in `afadaae`, confirmed by routing each file through the audit's own `vale-wrap.sh` (1 error each before, 0 after). Five further occurrences sit in `references/` files already on `main`; those are the pre-existing corpus and stay with #117, which is the real fix.

Also unfixed and not this PR's: `apm install` appends a duplicate `SessionStart` entry to `.claude/settings.json`, so a fresh clone cannot get pre-push green without an edit AGENTS.md warns against. Reproduces identically on `main`.

Co-authored-by: Defame1297 <gitea@rkdr.net>
Reviewed-on: https://git.dev.rkdr.net/Defame1297/holocron/pulls/129
Co-authored-by: Claude Code AI - Gitea MCP <claude@noreply.git.dev.rkdr.net>
Co-committed-by: Claude Code AI - Gitea MCP <claude@noreply.git.dev.rkdr.net>

plugins/lint/skills/vale-run/SKILL.md

@@ -1,28 +1,22 @@
 ---
 name: vale-run
 description: >
-  Use when running Vale (a prose/style linter) against files or directories in an
-  already-configured project — one that already has a .vale.ini — and interpreting
-  or reporting its results: choosing an output format for humans vs. CI, filtering
-  by severity, handling Vale's exit codes in scripts, or resolving common runtime
-  issues like false positives and unexpected CI failures. Use even if the user
-  doesn't say "vale" explicitly, e.g. "lint the docs", "check prose style", "run
-  the style linter", "why is CI failing on the docs check". Do not use when the
-  project has no .vale.ini yet, or needs styles installed/configured — that's the
-  vale-config skill.
+  Use when running Vale (a prose/style linter) on a project that already has a
+  .vale.ini and acting on its output — even when the user does not say "Vale",
+  as in "lint the docs", "check prose style", or "why is CI failing on the docs
+  check". Not setting up Vale config or styles -> `vale-config`.
 metadata:
-  version: "0.1.1"
+  version: "0.1.3"
   category: lint
   source_keys:
     - context7-websites-vale-sh
+    - house-vale-3-15-2-repro
 ---
 
 ## Gotchas
 
-- Vale's exit code is driven by `error`-level alerts only. `warning` and `suggestion` alerts are reported but still exit `0`. `MinAlertLevel` and `--minAlertLevel` control display, never the exit code — no flag makes warnings fail. A rule that must gate CI or a commit hook has to be `level: error`. This is the single most common way a Vale gate silently passes everything.
-- `vale ls-config` prints the fully-resolved, currently active configuration as JSON — the fastest way to check why a rule "isn't applying" is what's actually active, not what's written in `.vale.ini`.
-- Inline suppression syntax is format-specific: Markdown uses HTML comments `<!-- vale off -->` / `<!-- vale on -->`, MDX uses `{/* vale off */}` / `{/* vale on */}`, Org mode uses `# vale off` / `# vale on`. The MDX form does nothing in a plain `.md` file — the alert still fires. Don't assume one syntax works across formats.
-- Before calling the `vale` binary directly, check whether the target repo documents its own wrapper script for Vale (look in its README, CONTRIBUTING docs, pre-commit config, or a `scripts/` directory). Some projects wrap `vale` to work around real bugs — e.g. a scope that silently stops matching multi-line YAML block-scalar frontmatter fields — and calling bare `vale` in a repo that has such a wrapper silently skips whatever the wrapper works around. If a wrapper is documented, invoke it with the same arguments instead of calling `vale` directly; otherwise fall back to the default below.
+- Vale's exit code keys off `error`-level alerts only — `warning` and `suggestion` alerts print and still exit `0`, and `MinAlertLevel`/`--minAlertLevel` filter what is displayed, never the exit code — no flag makes warnings fail. A rule that must gate CI or a commit hook has to be `level: error`. This is the most common way a Vale gate silently passes everything.
+- Check whether the target repo documents its own `vale` wrapper script (README, CONTRIBUTING, pre-commit config, `scripts/`) before calling the binary. Some projects wrap `vale` to work around real bugs — e.g. a `text.frontmatter.<key>` scope that silently stops matching multi-line values (`>` folded scalars, plain continuation lines and quoted multi-line scalars all go unmatched; a `|` literal block scalar still works) — so bare `vale` skips whatever the wrapper fixes. Invoke the documented wrapper with the same arguments.
 
 ## Running vale
 
@@ -41,18 +35,22 @@ Key flags:
 | `--no-exit` | Suppresses the nonzero exit that `error`-level alerts would otherwise cause; a no-op when no rule is `error`-level. Use in CI stages that should surface lint output without hard-failing the build. |
 | `--ignore-syntax` | Treats input as plain text, skipping format-aware parsing — use when a file's syntax-aware parser produces noisy or wrong results. |
 
-`vale sync` downloads the packages/styles declared in `.vale.ini` — that's a one-time-per-change setup step (vale-config's territory), not part of a normal lint run. If a run behaves as though no styles are active, that's a sign `vale sync` hasn't been run yet, not a `vale-run` problem.
+`vale sync` downloads the packages/styles declared in `.vale.ini` — that's a one-time-per-change setup step (vale-config's territory), not part of a normal lint run. If a run behaves as though no styles are active, check `vale ls-config` before concluding it is a sync problem — an unrun `vale sync` is the usual cause, and not a `vale-run` problem. `ls-config` resolves config files, styles and `StylesPath` search paths; it never enumerates rules, so it cannot tell you a given rule is live.
 
 Prefer `--output=JSON` whenever the caller (a script, a CI step, another agent) needs to act on individual alerts rather than just get a pass/fail signal — `CLI` and `line` are for humans reading the terminal.
 
 ## Fixing false positives
 
+Before writing any inline suppression markup (steps 2 and 3 below), read `references/troubleshooting.md`: the form follows the parser the config picks, not the file extension, and the wrong form suppresses nothing while Vale reports no error.
+
+`.mdx` is the trap — the two parsers take opposite forms, so read `.vale.ini` first. Under `[formats] mdx = md` (what `vale-config` recommends) it is Markdown: `<!-- vale off -->` suppresses, `{/* vale off */}` does not. Without that mapping Vale takes the native MDX path, which needs the external `mdx2vast` binary (`npm install -g mdx2vast`); missing, the whole invocation dies — `E100 [lintMDX] ... mdx2vast not found`, exit 2 — leaving every other file in the run unlinted too.
+
 Scope the fix as narrowly as possible, in this order:
 
 1. **Mentioning banned phrasing rather than using it**: wrap it in backticks or a fenced code block. Vale skips code spans and fences, so no suppression is needed at all. Try this before any suppression markup.
 2. **One-off**: inline-suppress the specific text run with the format's `vale off`/`vale on` markup.
-3. **Recurring known-exception string, one rule**: disable that specific rule for that specific match inline (e.g. `<!-- vale Style.Redundancy["ACT test","OTHER"] = NO -->` ... `= YES`), rather than the whole rule.
-4. **Known project term failing spell check**: add it to the style's `ignore` list, not an inline suppression.
+3. **Recurring known-exception string, one rule**: disable that specific rule for that specific match inline (in Markdown, e.g. `<!-- vale Style.Redundancy["ACT test","OTHER"] = NO -->` ... `= YES`), rather than the whole rule.
+4. **Known project term failing spell check**: add it to the style's `ignore` list, not an inline suppression — and put the listed file at the `StylesPath` root, because an `ignore` path that resolves nowhere is a silent no-op (`references/troubleshooting.md`).
 
 Never disable a rule project-wide to fix one false positive — editing `.vale.ini`/`BasedOnStyles` is vale-config's job, and it silences the rule everywhere, not just the false-positive case.
 
@@ -60,4 +58,4 @@ If output looks wrong because Vale mis-parsed a file's format, rerun with `--ign
 
 For CI that fails solely because Vale returned non-zero on `error`-level alerts — not because the content is wrong for that pipeline stage — add `--no-exit` rather than disabling the rule. If the failing alerts are warnings or suggestions, Vale is not what failed the build; look elsewhere.
 
-If setting up Vale as a pre-commit hook or need the full inline-suppression/spelling-ignore syntax reference, read `references/troubleshooting.md`.
+If a rule appears not to apply, if you need the spelling-ignore syntax, if a CI failure still needs diagnosing, or if setting Vale up as a pre-commit hook, read `references/troubleshooting.md`.