docs: remind to update workflow meta-tests when changing CI workflows (#14239)

* docs: remind to update workflow meta-tests when changing CI workflows

#14205 moved Java and Desktop checks off pull_request/push triggers but
forgot three scripts/tests/*.test.js files that pin those workflows'
trigger shapes. The resulting red main CI run was fixed by #14227.
Add a step to the general workflow so the next workflow change updates
all parsing meta-tests, not just the obvious ones.

* docs: preserve workflow validation guidance after renumbering

Correct the self-audit reference so fixes rerun build, typecheck, and tests,
and keep the workflow lookup command on one line for Prettier.

* test(ci): update Java Maven step count for server Checkstyle

Co-authored-by: Qwen-Coder <[email protected]>

---------

Co-authored-by: qwen-code-dev-bot <[email protected]>
Co-authored-by: Qwen-Coder <[email protected]>

AGENTS.md

@@ -218,15 +218,21 @@ npm run preflight  # Full check: clean → install → format → lint → build
 2. **Test plan for behavioral changes** — write an E2E test plan in
    `.qwen/e2e-tests/` when the change affects user-observable behavior. Dry-run
    against the global `qwen` CLI first to confirm the baseline.
-3. **Build, typecheck, and test before declaring done**:
+3. **Update workflow meta-tests when changing CI definitions** — if you edit
+   `.github/workflows/*.yml`, run
+   `grep -R ".github/workflows/<filename>.yml" scripts/tests/` and update every
+   `scripts/tests/*.test.js` that parses the changed file. These tests pin
+   trigger shape and failure-routing contracts; leaving them stale turns a
+   workflow change into a red main CI run.
+4. **Build, typecheck, and test before declaring done**:
    `npm run build && npm run typecheck`, plus unit tests for the files you
    changed.
-4. **Self-audit before declaring done** — read the full diff you are about
+5. **Self-audit before declaring done** — read the full diff you are about
    to ship, including new untracked files, in open-ended passes, not hunting
    for anything specific. Then verify each change, and each green test you
    rely on as evidence, presuming it wrong (a passing test can assert the
    wrong thing). Stop after two consecutive clean passes — a clean pass is
-   evidence about that pass, not the code. A fix re-runs step 3, resets the
+   evidence about that pass, not the code. A fix re-runs step 4, resets the
    clean-pass count, and gets a further pass over the updated diff — never
    exit on a pass that found something. If five passes bring no convergence,
    say so instead of declaring done. Scale to the diff: one clean, careful