bump: switch to per-entry PR mode (one PR per stale plugin) (#2051)

* bump: switch to per-entry PR mode (one PR per stale plugin)

Replaces the single batched bump PR with one PR per stale plugin so a
single failing plugin no longer blocks the rest. Pins to a feature
branch of the bump-plugin-shas action that adds 'pr-mode: per-entry';
re-pin to the merge commit on the action's main when that lands.

- pr-mode: per-entry → one PR per plugin on bump/<slug>
- max_bumps default lowered 130 → 30 (per-entry scans cost more)
- scan dispatch fanned out over pr-urls JSON (one per per-entry branch)
- header comments updated for per-entry semantics

* bump: re-pin to merged composite action SHA on -community main

The pr-mode: per-entry input now lives on main of the bump-plugin-shas
action (merged at e2019b2a). Update the pin and drop the now-stale
header comment that tracked the feature branch.

* bump: dispatch all three required checks per per-entry PR

Bump PRs are opened with GITHUB_TOKEN, which doesn't fire on:pull_request
(recursion guard). The per-entry cutover already dispatched scan-plugins.yml
per branch to satisfy the `scan` required check, but `check` (Check MCP URLs)
and `validate` (Validate Plugins) are also required on main and likewise
never fired — leaving every bump PR BLOCKED on missing checks (observed on
the batched #2079, which only cleared after a human-authored push re-fired
the pull_request workflows).

Fix: dispatch all three workflows per per-entry bump branch. Each runs its
job unconditionally on workflow_dispatch, so the check run lands on the
branch HEAD (= PR head) and satisfies the required check.

- validate-plugins.yml: add workflow_dispatch trigger (check-mcp-urls.yml
  already had one). gh workflow run requires the trigger on the default
  branch; this lands together with the per-entry bump so main stays
  consistent.
- bump-plugin-shas.yml: loop the dispatch over
  {scan-plugins,check-mcp-urls,validate-plugins}; tolerate a single
  transient dispatch failure (warn, don't abort) so one hiccup can't
  strand the rest of the batch.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* bump: fail the per-entry check-dispatch step when a dispatch fails

The dispatch step logged each failed gh workflow run as a warning and exited 0, so a transient API error or rate limit could leave a per-entry bump PR missing a required check while the bump run still showed green. The composite action skips slugs with an open PR, so the stranded PR was never retried.

Attempt every dispatch (one failure must not strand the other branches), record failures via a temp file (the while loop runs in a pipe subshell), then emit an error and exit non-zero if any dispatch failed, so the bump run goes red and the affected PR can be re-dispatched.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

.github/workflows/bump-plugin-shas.yml

@@ -2,25 +2,24 @@ name: Bump Plugin SHAs
 
 # Nightly sweep: for each external entry whose upstream HEAD has moved past
 # its pinned SHA, validate at the new SHA with `claude plugin validate`
-# inline, then open one PR with all passing bumps. Each run force-resets the
-# bump/plugin-shas branch, so a previous night's unmerged PR is replaced (and
-# its review state discarded) — review and merge same-day to avoid churn.
+# inline, then open one PR per bumped plugin on branch `bump/<slug>`.
+# Failing entries stay isolated in their own PR; passing bumps merge
+# independently.
 #
 # Bot-free — uses the default GITHUB_TOKEN. PRs opened with GITHUB_TOKEN don't
-# trigger on:pull_request workflows, so the policy scan (`Scan Plugins`, a
-# required status check on main) would never run and the bump PR could never
-# merge. workflow_dispatch is exempt from that recursion guard, so we dispatch
-# the scan ourselves on the bump branch after the PR is opened. The check run
-# lands on the branch HEAD — the same SHA as the PR head — and satisfies the
-# required check.
+# trigger on:pull_request workflows, so the required status checks on main
+# (`scan` from Scan Plugins, `check` from Check MCP URLs, `validate` from
+# Validate Plugins) would never run and the bump PR could never merge.
+# workflow_dispatch is exempt from that recursion guard, so we dispatch all
+# three ourselves against each per-entry bump branch after its PR is opened.
+# Each check run lands on the branch HEAD — the same SHA as the PR head — and
+# satisfies the corresponding required check. (Each of those workflows runs
+# its job unconditionally on workflow_dispatch, so a dispatch always reports.)
 #
-# max-bumps is set above the external-entry count so a single run can clear
-# any backlog. The cost-control mechanisms are downstream:
-#   - scan-plugins.yml caches verdicts by (plugin, sha) so an unchanged SHA
-#     is never re-scanned across nightly force-resets.
-#   - revert-failed-bumps.yml drops policy-failing entries from the bump PR
-#     so one bad upstream can't block the rest.
-# See those files for details.
+# max-bumps caps the per-night work for cost control. Per-entry scans are
+# more expensive than a single batched scan, so the cap is conservative.
+# The composite action skips entries that already have an open bump PR, so
+# re-dispatches don't pile up duplicate work.
 
 on:
   schedule:
@@ -30,12 +29,12 @@ on:
       max_bumps:
         description: Cap on plugins bumped this run
         required: false
-        default: '130'
+        default: '30'
 
 permissions:
   contents: write
   pull-requests: write
-  actions: write  # gh workflow run scan-plugins.yml on the bump branch
+  actions: write  # gh workflow run {scan-plugins,check-mcp-urls,validate-plugins}.yml per bump branch
 
 concurrency:
   group: bump-plugin-shas
@@ -43,8 +42,8 @@ concurrency:
 jobs:
   bump:
     runs-on: ubuntu-latest
-    # Per-bump cost is ~2s (ls-remote + shallow clone + validate); 130 entries
-    # is ~5 min. The 60 min ceiling absorbs slow upstreams without letting a
+    # Per-bump cost is ~2s (ls-remote + shallow clone + validate); 30 entries
+    # is ~1-2 min. The 60 min ceiling absorbs slow upstreams without letting a
     # pathological run consume the default 360 min budget.
     timeout-minutes: 60
     steps:
@@ -52,18 +51,44 @@ jobs:
 
       # createCommitOnBranch-based bump so commits are signed by GitHub and
       # satisfy the org-level required_signatures ruleset on main.
-      - uses: anthropics/claude-plugins-community/.github/actions/bump-plugin-shas@c41c6911de0afffd2bc5cd8b21fb1e06444ee13b
+      - uses: anthropics/claude-plugins-community/.github/actions/bump-plugin-shas@e2019b2a01f11aa1484c53540b1cfab5eebbc299
         id: bump
         with:
           marketplace-path: .claude-plugin/marketplace.json
-          max-bumps: ${{ inputs.max_bumps || '130' }}
+          max-bumps: ${{ inputs.max_bumps || '30' }}
+          pr-mode: per-entry
           claude-cli-version: latest
 
-      # `bump/plugin-shas` is the action's default `pr-branch`. The scan diffs
-      # the branch against origin/main (the action's base-ref fallback when
-      # there's no pull_request event) and scans only the bumped entries.
-      - name: Dispatch policy scan on bump branch
-        if: steps.bump.outputs.pr-url != ''
+      # Per-entry fan-out: dispatch the three required checks against each bump
+      # branch. `pr-urls` is a JSON array of {name, old_sha, new_sha, branch,
+      # pr_url} entries emitted by the composite action when pr-mode is
+      # per-entry. All three (scan / check / validate) are required on main and
+      # none fire on the GITHUB_TOKEN-opened PR, so each must be dispatched.
+      # A single failed dispatch (transient API error / rate limit) must not
+      # strand the remaining branches, so we attempt every dispatch, then fail
+      # the step if any failed: a missing required check would otherwise leave
+      # its bump PR silently blocked behind a green run, and the composite
+      # action skips slugs with an open PR so it would never be retried.
+      - name: Dispatch required checks per per-entry PR
+        if: steps.bump.outputs.pr-urls != '' && steps.bump.outputs.pr-urls != '[]'
         env:
           GH_TOKEN: ${{ github.token }}
-        run: gh workflow run scan-plugins.yml --ref bump/plugin-shas
+          PR_URLS: ${{ steps.bump.outputs.pr-urls }}
+        run: |
+          set -euo pipefail
+          dispatch_failures="$(mktemp)"
+          jq -c '.[]' <<<"$PR_URLS" | while read -r entry; do
+            branch=$(jq -r '.branch' <<<"$entry")
+            name=$(jq -r '.name' <<<"$entry")
+            for wf in scan-plugins check-mcp-urls validate-plugins; do
+              echo "Dispatching ${wf}.yml against $branch ($name)"
+              if ! gh workflow run "${wf}.yml" --ref "$branch"; then
+                echo "::error::Failed to dispatch ${wf}.yml against $branch ($name) — required check will be missing; re-dispatch with: gh workflow run ${wf}.yml --ref $branch"
+                echo "${wf} ${branch}" >> "$dispatch_failures"
+              fi
+            done
+          done
+          if [ -s "$dispatch_failures" ]; then
+            echo "::error::$(wc -l < "$dispatch_failures" | tr -d ' ') required-check dispatch(es) failed; the affected bump PR(s) are blocked until re-dispatched (see annotations above)."
+            exit 1
+          fi

.github/workflows/validate-plugins.yml

@@ -12,6 +12,14 @@ on:
     branches: [main]
     paths:
       - '.claude-plugin/**'
+  # `validate` is a required status check on main. Bump PRs are opened with
+  # GITHUB_TOKEN, which doesn't fire on:pull_request (recursion guard), so the
+  # path-filtered trigger above never reports on them and the PR would be
+  # blocked forever. The bump workflow dispatches this against each per-entry
+  # bump branch instead; the check run lands on the branch HEAD (= PR head)
+  # and satisfies the required check. The validate job runs unconditionally,
+  # so a dispatch always reports.
+  workflow_dispatch:
 
 permissions:
   contents: read