Add an additional answer step at the end of orchestration finish

[lolipopshock]

for example, #372

it's very unclear what happens / what's implemented at the end of a long orchestrated execution. in that case, we should just add another answer step and summarize what's implemented ? and the progres?

[lolipopshock]

@sepo-agent /answer

[sepo-agent-app]

Yes, the product need is real. I would not implement it as “dispatch another /answer action” though. I’d add a first-class orchestration closeout comment at the terminal point of the orchestrator.

Right now, orchestrated runs hand back after post-processing, then either dispatch another workflow or stop. The relevant surfaces are:

• Orchestrated workflows hand back only when orchestration_enabled: true: https://github.com/self-evolving/repo/blob/main/.agent/docs/architecture/agent-orchestrator.md#L57-L65
• Review dispatches back into the orchestrator after posting synthesis: https://github.com/self-evolving/repo/blob/main/.github/workflows/agent-review.yml#L379-L402
• Terminal stop comments already exist, but are mostly structural: https://github.com/self-evolving/repo/blob/main/.agent/src/cli/orchestrate-handoff.ts#L1148-L1178
• Ordinary terminal PR stops are explicitly silent today, which explains the PR Harden issue-body auth association fallback #372 confusion: https://github.com/self-evolving/repo/blob/main/.agent/docs/architecture/agent-orchestrator.md#L145-L152

Recommended approach:

Add an orchestrator-owned final summary comment when a chain reaches a terminal state, especially after review returns SHIP / NO_AUTOMATED_ACTION, self-approval stops, self-merge stops, or the round budget is exhausted. The comment should summarize:

• What was implemented or changed
• PR(s) created/updated
• Review/fix/review progress across rounds
• Latest verification/test status when available
• Final state, for example SHIP, NEEDS HUMAN DECISION, STOPPED, or BLOCKED
• Recommended next human action

This should be generated from existing artifacts: implementation PR body, fix-pr status comments, latest review synthesis, handoff markers, and source conclusion. That keeps it deterministic and avoids spawning a loose /answer run that has to rediscover state.

Execution plan:

1. Add a closeout formatter in orchestrate-handoff.ts next to formatOrchestrateStopComment.
2. Trigger it for terminal orchestrated PR stops, not just parent issue/meta stops.
3. Reuse existing dedupe marker behavior with a new hidden marker, e.g. sepo-agent-orchestrate-closeout.
4. Gather latest relevant status from PR comments/review synthesis using existing GitHub helper patterns.
5. Keep the summary compact: final state first, then progress bullets, then next action.
6. Update orchestrator docs to say terminal orchestration posts a final closeout summary.
7. Add tests in orchestrate-handoff-cli.test.ts covering a PR chain like Harden issue-body auth association fallback #372 ending in SHIP / NO_AUTOMATED_ACTION.

So: yes, add the end summary, but make it an orchestrator closeout/reporting step rather than a new /answer dispatch.

---

codex | gpt-5.5/medium | high | Macmini-runner-3

[github-actions]

Closing because this agent issue has had no activity for 30 days.