chore: update fullsend shim workflow

[fullsend-ai-fullsend]

This PR updates the fullsend shim workflow to match the current template in the .fullsend config repo.

The shim content has drifted from the template — this brings it back in sync.

[github-actions]

E2E tests did not run

E2E tests run automatically for org/repo members and collaborators on pull requests.

For other contributors, a maintainer must add the ok-to-test label after the latest push.

See E2E testing guide for details.

[github-actions]

Site preview

Preview: https://99f2932f-site.fullsend-ai.workers.dev

Commit: c587cb06983d37ac0940e3133aaa1f2a1520c522

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 6:17 AM UTC · Completed 6:30 AM UTC
Commit: c587cb0 · View workflow run →

[fullsend-ai-retro]

Retro: PR #5559 — chore: update fullsend shim workflow

Workflow outcome: Clean and efficient. Bot-authored template sync PR merged in 17 minutes with one human approval.

Timeline

1. 05:58 UTC — fullsend-ai-fullsend[bot] created PR on fullsend/onboard branch, updating .github/workflows/fullsend.yaml to skip dispatch for fullsend/scaffold-install branches
2. ~06:00 UTC — Initial CI run (30070903781) failed — TestManagedShimStopFixNotStale caught stale stop-fix authorization patterns in the rendered shim (related to recently closed #5421)
3. 06:07 UTC — Bot force-pushed corrected content, bringing the rendered shim into sync with the updated template
4. 06:09 UTC — rh-hemartin approved; auto-merge enabled
5. 06:10 UTC — Added to merge queue; re-run CI (30071335272) passed
6. 06:15 UTC — Merged

Key observations

• Zero agent involvement: No review, code, or fix agents dispatched. The only agent run was this retro itself.
• CI failure → force-push pattern: The initial push used a template version that predated the dispatch: stop-fix shim job still uses author_association instead of collaborator permission API #5421 stop-fix authorization fix (merged Jul 23). The TestManagedShimStopFixNotStale staleness guard caught this correctly, and the force-push resolved it. This wasted one CI run.
• Retro dispatch on zero-agent PR: This retro agent consumed an opus-model session to analyze a PR with no agent workflow to examine.

Existing issue coverage

All improvement areas identified are already tracked by open issues:

• #5229: Skip retro on reconciliation branch PRs — this PR is another instance of the exact pattern described (automated fullsend/onboard PR triggering a full retro run with nothing to analyze).
• #5295 / #5399: Skip retro for bot-authored PRs with zero agent involvement — also directly applicable.
• #1244 / #2574: Pre-validate reconciliation shim content before creating PRs — the CI-failure-then-force-push pattern adds evidence that the reconciliation bot would benefit from pre-push validation against the test suite, not just actionlint.

No new proposals filed — existing issues cover all findings.