cc-release-watch → claude-release-watch

Why renaming the workflow file is the right escape hatch when GHA's metadata cache is wedged.

What we tried first

Why content edits don't unstick the cache

GitHub Actions associates each workflow with a numeric ID (here, 259675442) at first registration. Subsequent edits to the file SHOULD trigger a re-parse — but in this case, the workflow's metadata (name: and recognized triggers) has been frozen since the file was first added on 2026-04-12, regardless of how many times the contents change.
Possible causes (we don't need to know which to fix it): a partial registration on first push, a cache invalidation regression on GHA's side, or an internal state we can't observe. What matters: edits to file content don't fix it.

The fix: rename to force a new workflow ID

Before

.github/workflows/cc-release-watch.yml
  name: cc-release-watch
  workflow ID: 259675442 (cached, broken)

After

.github/workflows/claude-release-watch.yml
  name: claude-release-watch
  workflow ID: (new — clean metadata)
GHA assigns a fresh workflow ID per unique file path. Renaming the file forces a clean registration with the current on: block, including workflow_dispatch: {}.

Tradeoff

Verification path

  1. This PR merges to main.
  2. gh workflow list shows claude-release-watch as a friendly name (not file path).
  3. gh workflow run claude-release-watch.yml returns a run URL — no more 422.
  4. The dispatched run executes end-to-end, including the LLM triage step (now that CLAUDE_CODE_OAUTH_TOKEN is configured), and produces either zero feature gaps (no new CC versions today) or new cc-adoption issues (if any high-score gaps exist).

Followup

Once the new workflow has fired successfully at least once, we can gh api -X DELETE the orphaned ID 259675442 to keep gh workflow list --all clean. Optional cosmetic step; doesn't affect correctness.