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/gitea/.apm/skills/gitea-files/SKILL.md

@@ -2,15 +2,9 @@
 name: gitea-files
 
 description: >
-  Use when reading or writing individual files or directory trees in a Gitea repository via the
-  Gitea MCP server: reading a file's contents, listing a directory, walking a full repository
-  tree, creating a new file, updating an existing file, or deleting a file. Triggers on "read this
-  file from the repo", "what's in this directory", "show me the repo tree", "create/update a file
-  in Gitea", "commit this file to the branch", "delete this file from the repo" — even when the
-  user doesn't say "Gitea" explicitly, as long as the target is a Gitea-hosted repository. Do not
-  use for local filesystem file operations (use Read/Write/Edit), for branch or commit history
-  (use gitea-branches), or for opening a pull request around a file change (use gitea-prs after
-  the file write completes here).
+  Use when reading or writing files or directories in a Gitea repository via the MCP server,
+  rather than the local filesystem (for a local path use Read/Write/Edit) — even when the user
+  does not say "Gitea". Not commit history -> `gitea-branches`. Not pull requests -> `gitea-prs`.
 
 compatibility: Requires the Gitea MCP server configured with a token scoped to at least
   write:repository. Tested with a token holding write:issue + write:repository; write:issue
@@ -28,29 +22,25 @@ allowed-tools: mcp__gitea__get_file_contents mcp__gitea__get_dir_contents mcp__g
 
 ## Gotchas
 
-- **A 404 from any read call may actually be a 403 in disguise.** `get_file_contents`, `get_dir_contents`, and `get_repository_tree` all gate on `write:repository` scope, not just read access — some Gitea endpoints return 404 instead of 403 when the token's scope is insufficient, to avoid leaking whether the resource exists. If a read fails with 404 on a path you're confident is correct, check the token's configured scopes before concluding the file or directory doesn't exist.
-- **SHA is the concurrency token for every write — and it lives at the top level of `get_file_contents`'s response, not nested under `content`.** `create_or_update_file` without `sha` is always treated as a *create*: if the path already exists, Gitea returns HTTP 409. `delete_file` has no optional path at all — omitting `sha` returns HTTP 422. The safe sequence for any update or delete is always: call `get_file_contents` first, read the top-level `sha` field, then pass that exact value to the write call. Never guess or reuse a stale SHA — a mismatched SHA is rejected the same as a missing one.
-- **A write can also fail because the branch requires signed commits — a separate failure mode from a bad SHA.** `create_or_update_file` and `delete_file` create commits server-side via a bare API token call with no 2FA/PGP context. If the target branch's protection rule requires signed commits, Gitea rejects the write outright — surfaced as a generic 403 or 422, not an error naming "signed commit required," and reads against that same branch keep succeeding right up until you try to write. When a write fails without a clean 409 (missing/stale SHA) or 404 (bad path) explanation, check whether the branch's protection rule requires signed commits before assuming the SHA is wrong and retrying.
-- **A large `create_or_update_file` payload can hit a reverse-proxy 413 that has nothing to do with Gitea.** `content` is base64-encoded, which inflates the payload ~33% over the raw file size; a 413 is commonly a reverse-proxy body-size limit in front of the Gitea instance, not a Gitea-side rejection. No amount of retrying, or changing the SHA, path, or branch, will fix it — it needs the proxy's config raised, which is outside this skill's or the calling agent's control. Surface that distinction to the user instead of retrying the same call.
-- **`get_dir_contents` and `get_repository_tree` are not SHA sources for a specific file's write.** `get_dir_contents` entries carry no `sha` at all. `get_repository_tree` entries do carry a `sha` (a blob/tree hash), but fetching it means an extra round trip with no content — `get_file_contents` is the canonical path since it returns the decoded content and the write-ready `sha` in one call.
-- **`owner` and `repo` are always caller-supplied inputs, never resolved here.** This skill doesn't infer them from a git remote. If invoked directly by a human, ask for them if not stated. If invoked by `gitea-workflow` or an orchestrating agent, expect them to already be resolved and passed in.
-- **Direct commits to a branch are a first-class action, not a workaround.** Gitea's own web UI defaults to editing files directly against a branch — `create_or_update_file`/`delete_file` used that way is normal, not an API escape hatch to avoid. The SHA-currency requirement above is the actual risk to manage, not the act of committing directly.
-- **`ref` (reads) vs. `branch_name` (writes) are different parameters for the same concept.** `get_file_contents`, `get_dir_contents`, and `get_repository_tree` (as `tree_sha`) all accept a branch name, tag, or commit SHA to select what to read. `create_or_update_file` and `delete_file` instead take `branch_name` — the branch the commit lands on. Don't conflate the two when chaining a read into a write.
-- **Content is base64.** `create_or_update_file`'s `content` parameter is base64-encoded file content, not raw text — encode before calling. `get_file_contents`'s response content is likewise base64-encoded (decode after reading), unless `withLines: true` is passed for a numbered-line view.
+- **A 404 may mean an under-scoped token, not a missing path.** Every tool here gates on `write:repository`, and Gitea masks insufficient scope as 404. Check scopes first.
+- **Reads take `ref` (`tree_sha` on `get_repository_tree`), writes take `branch_name`.** One concept, three names — carry the wrong key and the branch is dropped.
+- **`content` is base64 both ways — except under `withLines: true`.** Encode before a write, decode after a read; but with `withLines: true` `content` is already plain JSON text and the reported `"encoding": "base64"` is a lie. Decoding it yields garbage.
 
-## Reading
+## Inputs
 
-- **Single file:** `get_file_contents(owner, repo, ref, path)`. Pass `withLines: true` only when you need line numbers for referencing specific lines (e.g. quoting a snippet back to the user); omit it for a normal content fetch.
-- **One directory level:** `get_dir_contents(owner, repo, ref, path)` — returns immediate entries only (name, path, type, size), no recursion, no SHA, no content.
-- **Whole tree:** `get_repository_tree(owner, repo, tree_sha, recursive)` — `tree_sha` accepts a SHA, branch, or tag name despite the name. Set `recursive: true` to walk subdirectories in one call. Response includes `truncated: true` when a page doesn't hold every entry — page through with `page`/`per_page` (default `page: 1`, `per_page: 30`) until you get fewer results than `per_page`.
+`owner`, `repo` and the target branch are caller-supplied. This skill never infers them from a git remote: ask the human when they are not stated, and expect an orchestrating caller to have resolved them already.
 
-## Writing
+## Dispatch
 
-- **Creating a new file:** call `create_or_update_file(owner, repo, path, content, message, branch_name)` with `sha` omitted entirely.
-- **Updating an existing file:** call `get_file_contents(owner, repo, ref: branch_name, path)` first, take the top-level `sha`, then call `create_or_update_file(..., sha: <that value>)`.
-- **Deleting a file:** call `get_file_contents` first the same way, then `delete_file(owner, repo, path, message, branch_name, sha: <that value>)` — `sha` is required, no create-style fallback exists.
-- **Creating a new branch as part of the write:** pass `new_branch_name` on `create_or_update_file` to branch off before the commit lands, instead of calling a separate branch-creation step.
+Read the reference for the row you land on before making the call.
 
-If the change needs review before merging, or targets a protected branch, hand off to `gitea-prs` after the write lands here to open the pull request — this skill's scope ends at the commit.
+| Condition | Flow | Reference |
+|---|---|---|
+| Read one file, list one directory level, or walk the repository tree | Read | `references/reading.md` — the three read tools, their pagination behaviour, and why neither a directory listing nor a tree entry supplies the SHA a write needs |
+| Create, update, or delete a file | Write | `references/writing.md` — the SHA-first sequence every update and delete depends on, the worked multi-call sequence, and how to triage a write that fails |
 
-If you need the full multi-call sequence rather than the single-call summary above — e.g. branching off as part of a file push ahead of opening a PR, or recovering a SHA you didn't capture earlier — read `references/examples.md`.
+A request that reads and then writes runs both flows in that order: fetch the file first, then write with the SHA that call returned.
+
+## Handoff
+
+Scope ends at the commit. Gitea's own web UI edits files directly against a branch, so committing straight to a branch is the normal path rather than an escape hatch — hand off to `gitea-prs` when the change needs review before merging or the target branch is protected, not by default.