Skip fullsend review dispatch for changesets-release branches

[fullsend-ai-retro]

What happened

PR #4428 is an automated changesets version-bump PR on branch changesets-release/ai-integrations/main. The review agent ran (workflow 32422250854), spent ~7 minutes, and approved with a trivial "Looks good to me" — no inline comments, no findings. The diff was 62 additions / 20 deletions across 17 files, all mechanical: package.json version bumps and auto-generated CHANGELOG.md entries. The underlying changes were already reviewed when the originating code PRs merged their changesets. This pattern repeated on 11 of 20 recent merged changeset PRs (55%) across the 3 workspaces in the paths filter. PR #4428 accumulated 8 review dispatches over its lifetime as the long-lived branch was repeatedly synchronized.

What could go better

The review agent's token and compute budget is spent reviewing changes that are (a) purely mechanical version-number edits and changelog boilerplate, and (b) already reviewed as part of the original code PRs that created the changesets. Every review produces a trivial LGTM. The current fullsend.yaml workflow has no condition to distinguish changesets-release branches from code branches. Note: issue #4166 shows the team wants bot-authored PRs reviewed in general — but changesets version-bump PRs are a distinct mechanical subset where full review adds near-zero value compared to, e.g., Renovate dependency bumps that could introduce breaking changes. Confidence: high — the evidence across 20 PRs is consistent (zero non-trivial findings on any changeset PR).

Proposed change

Add a branch-name filter to the pull_request_target trigger in .github/workflows/fullsend.yaml to exclude changesets-release/* branches from dispatching the review (and retro) agents. Specifically, add a branches-ignore clause or an if condition on the dispatch jobs:

pull_request_target:
  types: [opened, synchronize, ready_for_review, closed]
  branches-ignore:
    - 'changesets-release/**'
  paths:
    - "workspaces/boost/**"
    - "workspaces/scorecard/**"
    - "workspaces/ai-integrations/**"

Alternatively, if the team prefers to keep the workflow triggered (e.g., for future use) but skip agent dispatch, add an if condition on the review/retro job steps: if: "!startsWith(github.event.pull_request.head.ref, 'changesets-release/')". The branches-ignore approach is simpler and prevents the workflow from running at all on changeset branches.

Validation criteria

After the change, the next 5 merged "Version Packages" PRs in workspaces covered by the paths filter (boost, scorecard, ai-integrations) should have zero fullsend-ai-review[bot] comments and zero review workflow dispatches. Verify by checking gh run list --repo redhat-developer/rhdh-plugins --branch changesets-release/<workspace>/main for absence of new fullsend runs.

---

Generated by retro agent from #4428

[fullsend-ai-triage]

🤖 Finished Triage · ✅ Success · Started 10:04 PM UTC · Completed 10:09 PM UTC

Commit: db7bee2 · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $0.78

[fullsend-ai-triage]

⚠ Workflow change required — manual intervention needed. The code agent cannot modify workflow files under current permissions.

Implementation notes

The proposed branches-ignore approach will not work here. For pull_request_target, the branches / branches-ignore filter acts on the base branch (i.e., main), not the PR's head branch (changesets-release/*). The correct approach is the if-condition alternative:

dispatch:
  if: >-
    (github.event_name != 'pull_request_target'
     || !startsWith(github.event.pull_request.head.ref, 'changesets-release/'))
    && (github.event_name != 'issue_comment'
        || github.event.comment.user.type != 'Bot')

Additionally, .github/workflows/fullsend.yaml is managed by the fullsend upstream scaffold and states "Do not edit it directly." A local edit would be overwritten on the next scaffold sync. The fix likely needs upstream support in fullsend-ai/fullsend (e.g., a configurable head_branches_ignore parameter in .fullsend/config.yaml), or needs to be applied in the reusable-dispatch.yml layer.

Related issues

• Issue Update fullsend.yaml workflow SHA to restore review dispatch for bot-authored PRs #4166 (open) establishes the policy that bot-authored PRs should be reviewed — this issue correctly carves out a narrower exception for mechanical changesets PRs, which is consistent with that policy.

Proposed test

# After fix: confirm no new fullsend runs on changesets-release branches
gh run list --repo redhat-developer/rhdh-plugins \
  --branch changesets-release/ai-integrations/main \
  --workflow=fullsend.yaml --limit 5
# Should show no runs after the fix date

# Confirm normal code PRs still trigger review
gh run list --repo redhat-developer/rhdh-plugins \
  --workflow=fullsend.yaml --limit 5
# Should show recent runs for non-changesets PRs

---

Labels: Issue targets the fullsend workflow configuration and proposes a new filtering feature to reduce unnecessary agent dispatches.

---

Next steps:

• /fs-code — agent creates a PR to implement this issue
• /fs-code <your instruction> — agent implements with your specific guidance