Does anything need us?

Reference: CONTRIBUTING.md (contributor PRs, stacked drafts, "Depends on #N"), docs/dev/RELEASE_PROCESS.md (the draft-only release workflow and the Foundry package-release endpoint), deploy/e2e-runner/README.md and the header comment of .github/workflows/e2e-gate.yml (who may run E2E), the header comments of audit-fix.yml and dependabot-auto-merge.yml (what the bot PRs expect). This skill is the checklist; those are the truth when the two disagree.

Read-only. Report and recommend; never label, comment, close, merge, rebase, delete a branch, or run a workflow as part of the check. The only thing the scan changes is the remote-tracking refs (git fetch --prune); it never touches a local branch, the working tree, or GitHub. Never publish or promote a release — the maintainer does that by hand, always (see the standing authorizations in CLAUDE.md). Offer the action and wait for a yes.

1. Scan:

   .claude/skills/needanything/scan.sh
   

   About 20 seconds; one tab-separated line per item. It reads foundryvtt-dcc/dcc, the lib repo moonloch/dcc-core-lib, the sibling modules (xcc, mcc-classes, dcc-crawl-classes, dcc-core-book — dcc-qol is deliberately skipped), and the local checkout. gh output is paged off with GH_PAGER=cat; the script sets it, so don't drop that if you edit it.

2. Classify each open PR (section OPEN PRS). There is no issue automation in this repo: every PR was opened by Tim, a contributor, or one of two bots.

   PR state	Meaning / suggested action
   author=app/dependabot, checks pending	In flight — dependabot-auto-merge.yml squash-merges it as soon as Run Tests succeeds on its current head (main has no required checks on purpose: a required check would also reject the release workflow's push). Say nothing unless it has sat for more than a day.
   author=app/dependabot, Run Tests green, still open	The merge step skipped it: conflicts with main (merge=DIRTY, wait for Dependabot's rebase or resolve it by hand and push — the next green run merges it), a draft, or a push newer than the run. Needs a human only if it stays that way. Red Run Tests on it → read the failure first.
   author=app/github-actions, branch automated/npm-audit-fix	The daily Audit Fix PR. Run Tests never runs on it (opened with GITHUB_TOKEN); the unit tests ran inside the workflow before it opened, so no checks on it is normal (merge=UNSTABLE or CLEAN; BLOCKED if Run Tests is ever made required, which then needs an admin merge). Needs you: read the lockfile diff and merge. It re-opens or updates itself daily until merged.
   author is not Tim and not a bot	Contributor PR. Review per CONTRIBUTING.md: ask for specific changes if needed, then merge as submitted — never cherry-pick or trim it on their behalf. Say how long it has waited.
   e2e=PENDING (or -) with no e2e-approved label, author not a member	The E2E gate is waiting: either Claude triage is still running, or it declined and a maintainer must add the e2e-approved label to run the PR's code on the self-hosted runner. A new push drops the label again.
   e2e=FAILURE	Either the gate's Claude triage declined the PR (the gate's PR comment says so; a maintainer overrides with e2e-approved) or the suite is red on the self-hosted runner. Read the status link to tell which. For a red suite, check the log before blaming the change — some specs are order-dependent or scene-drift flaky (docs/dev/TESTING.md); a standalone rerun is the first step.
   e2e=ERROR	The gate itself crashed (triage error), so nothing ran. A maintainer adds e2e-approved or the author re-pushes to retry.
   checks_failed non-empty	Red CI — name the failing check(s) (Run Tests, Check Translations, CodeQL, …).
   merge=DIRTY (or CONFLICTING)	Conflicts with main — needs a rebase. A contributor's PR: ask them; ours: offer to do it.
   draft=true, depends_on=#N(MERGED)	Stacked PR whose base has landed. It should be rebased onto main and marked ready (the author's job for a fork; ours if it's our branch). Offer the nudge or the rebase.
   draft=true, depends_on=#N(OPEN)	Waiting on #N — no action.
   draft=true, no dependency	Work in progress — no action unless it's older than a week, then ask whether it's abandoned.
   merge=BLOCKED, checks green	Waiting on a required review/approval.
   ready, green, merge=CLEAN, author Tim or a contributor	Green but unmerged — waiting on Tim's review/merge. After it lands, /release is the usual next step.
   checks_pending > 0, or merge=UNKNOWN on a fresh push	In flight — not yet actionable.

   A PR updated in the last few minutes may still be settling; say so rather than calling it stuck.

3. Issues. There is no triage pipeline; the open issues are Tim's own backlog unless someone else shows up in them.

   • ISSUES AWAITING A REPLY — someone other than Tim spoke last (or opened it and nobody answered). Needs you: read the newest comments (gh issue view <n> --comments) and say what they're asking.
   • BUG / FOUNDRY-API-CHANGE ISSUES — list these ahead of enhancements; a foundry-api-change issue usually means a Foundry update broke something and players are seeing it.
   • BACKLOG — the rest is available to pick up; give the count and the lowest number. Don't list them one by one.

4. Release state (section RELEASE).

   • draft_release= line → Needs you now: a draft is waiting to be published, and only the maintainer may publish it (CLAUDE.md standing authorizations). Offer to tidy its notes, never to promote it.
   • unreleased_commits > 0 and no draft → feature work sits on main unreleased. Offer the version.txt bump from /release (its own commit on main, never inside a PR). CI-only or docs-only commits don't need a release; say so instead of offering one.
   • bump_unreleased=yes with no draft and no Create GitHub Release run newer than the bump → the release workflow was skipped. A very large squash-merge can trip the paths: version.txt filter; recovery is a small whitespace push to version.txt on main, which needs Tim's go-ahead.
   • Create GitHub Release red → no draft was made; read gh run view <run> --log-failed.
   • Foundry Website Update or Update Foundry Manifest Post-Release red → a release was published but Foundry's registry never heard of it: in-app installs stay frozen on the previous version. Needs you now. See docs/dev/RELEASE_PROCESS.md § Manifest URLs & the package-release endpoint.

5. Scheduled and push jobs on main (NEWEST RUN PER WORKFLOW ON MAIN, plus one red_now= line per workflow whose newest run failed). Green → say nothing. Red:

   • Run Tests on main → main is red; name the commit.
   • Audit Fix → the daily audit either found an advisory with no in-range fix (left for a human) or the job itself broke at Create or update fix PR; read the log and say which.
   • Dependabot Updates → Dependabot can't produce its bumps; a long red streak here means security updates silently stopped. Read the log, file an issue if none tracks it.
   • CodeQL → read the alert; usually a real finding or a config break.

6. Core lib (CORE LIB).

   • sync_due=yes (vendored behind the latest published lib release) → a /sync-core-lib is due. Note whether the branch carries a temporary adapter workaround that the sync should remove.
   • Our own open lib PR → after it merges, the sync + workaround removal is the next step; a lib Dependabot PR with green checks is a one-click merge for Tim.
   • Lib issues awaiting a reply → someone outside is asking; the lib is public and must stay free of book text, so check any answer for that.

7. Sibling modules (SIBLING MODULES): open PRs and issues awaiting a reply in xcc, mcc-classes, dcc-crawl-classes and dcc-core-book. These are usually low priority; list them under Worth a look with the module name and age, and flag a module PR that depends on an unreleased DCC change.

8. Local checkout (LOCAL CHECKOUT): housekeeping, never urgent.

   • dirty_files on main → say what's uncommitted before anyone switches branches (Tim does his own git work in this checkout).
   • fetch=failed → the ahead/gone lines are stale; say so.
   • branch=… [ahead N] → unpushed commits; offer the push if the branch isn't main.
   • gone_branches= → local branches whose remote was deleted. PRs are squash-merged, so git branch -d refuses them; before offering git branch -D, confirm each one's PR merged (gh pr list --head <branch> --state merged) and leave out any that didn't.
   • A worktree whose PR: #N MERGED → offer pnpm run work:finish.

9. Report, grouped, most urgent first — one line per item with the number, the reason, and the one next step:

   Needs you now
   - Draft release v0.70.76 waiting 3h — publish it (yours to do)
   - PR #1094 — Audit Fix lockfile bump, no checks by design; review and merge
   - #1101 — reporter replied 2h ago with a repro; they're waiting on an answer
   
   Worth a look
   - Dependabot Updates red 6 runs in a row on main (Run Dependabot step) — no issue tracks it
   - mcc-classes #30 — PhDoom's sheet request, 632d without a reply
   - 8 local branches with a deleted remote (all merged PRs) to delete; worktree for #1075 can be finished
   
   In flight (no action)
   - PR #1102 — contributor PR, E2E gate triage running
   

   If nothing qualifies, say so in one line and give the counts checked (for example: "Nothing needs you — 1 open PR (bot, in flight), 0 issues awaiting a reply, 23 backlog issues, v0.70.75 published with nothing unreleased, main green, lib in sync."). Skip empty groups. Finish by offering the obvious next action (merge the audit-fix PR, run /release, start /sync-core-lib, pick up the lowest-numbered backlog issue) and do it only on a yes.