agent-pr-autoloop

name	agent-pr-autoloop
version	1.0.0
description	Run LoRAIro PR maintenance to completion automatically after an agent creates a PR or a draft PR becomes ready for review: keep polling CI and bot review with gh, repair failures and reply in Japanese in the same worktree, escalate design loops, and squash merge when safe, without waiting for a human to restart each poll cycle. Use right after an agent-created PR exists or transitions draft-to-ready. Do NOT use for human-authored PRs, and do NOT redefine repair/merge/escalation policy here (that lives in agent-pr-maintainer).
metadata	{"short-description":"PR作成後の保守ループを人手の再起動なしに最後まで自走させる共通スキル。作法差分はCLAUDE.md/AGENTS.mdで吸収。"}
dependencies	["agent-pr-maintainer","github-ops"]

Agent PR Autoloop

Automatically drive the agent PR maintenance loop to a terminal state after a PR is created, without a human restarting each poll. This skill is shared by Codex and Claude Code; only the non-blocking wait mechanism differs per agent and is absorbed by each agent guide (see "Per-Agent Wait Mechanism").

Relationship to agent-pr-maintainer

This skill is the autonomous looping layer. It does not define policy.

• All policy — draft gates, repair rules, repair-loop limit (4), Japanese review replies, design escalation, merge conditions, post-merge worktree cleanup — comes from the agent-pr-maintainer skill and ADR 0039.
• This skill only specifies the contract for how to keep the loop running by itself until a terminal state.

If anything here appears to conflict with agent-pr-maintainer, agent-pr-maintainer wins on policy and this skill wins only on "do not stall waiting for a human between polls".

When to Use

Use immediately when any of these is true:

• This agent just created a LoRAIro PR (gh pr create), including a draft created only because workflow says so.
• A draft agent PR transitioned to ready for review (including a manual transition by the user).
• The user asks to continue/finish an agent-created PR through CI, review, repair, and merge.

First enforce the no-draft postcondition from agent-pr-maintainer (ready the PR if it is reviewable), then enter the loop. Do not stop at the PR URL when the PR is reviewable.

Autoloop Contract (agent-agnostic)

Treat maintenance as a self-paced poll loop. One loop cycle:

1. Gather state with gh — all four checks are mandatory every cycle; never skip based on CI status:

   # (a) PR fields — isDraft, mergeStateStatus, reviewDecision, head SHA
   gh pr view "$PR" --repo "$REPO" \
     --json number,isDraft,headRefOid,mergeStateStatus,reviewDecision,statusCheckRollup,labels
   
   # (b) Check run statuses
   gh pr checks "$PR" --repo "$REPO" \
     --json name,state,bucket,link,startedAt,completedAt,workflow
   
   # (c) Bot reactions on the PR issue (Codex signals: +1=clean, eyes=in-progress)
   gh api "repos/$REPO/issues/$PR_NUM/reactions"
   
   # (d) Inline review comments — detect P1/P2 badge findings from Codex
   gh api "repos/$REPO/pulls/$PR_NUM/comments"
   

   $REPO is owner/repo (e.g. NEXTAltair/LoRAIro or NEXTAltair/genai-tag-db-tools). mergeStateStatus=CLEAN means GitHub merge preconditions pass; it does not mean bot review is complete.

2. Classify the cycle into exactly one outcome:

   • continue — any of these: CI checks still running (step b has pending checks); no chatgpt-codex-connector[bot] reaction yet on the PR issue (step c is empty or only eyes); CI is green but Codex review/reaction has not posted yet → schedule the next cycle.

   • repair — any of these: CI checks failed (step b); Codex review state is COMMENTED and step d contains P1/P2 badge findings → fix in the PR worktree, validate, push, reply in Japanese per agent-pr-maintainer, then continue (count the repair against the limit).

   • escalate — repair limit reached, recurring findings in the same boundary, or a forbidden change is required → open the design issue, comment in Japanese, stop.

   • merge — all of the following are true (verified from step 1 data above):

     1. CI: all required checks pass (step b — no failed or pending).
     2. Draft: isDraft == false (step a).
     3. Codex reaction: chatgpt-codex-connector[bot] has a +1 reaction on the PR issue (step c). eyes alone → continue, not merge.
     4. No blocking findings: step d has no unaddressed P1/P2 Codex comments.

     ⚠️ mergeStateStatus=CLEAN with reviewDecision="" alone does not satisfy this gate. → squash merge, clean up the worktree, stop.

   • timeout — total elapsed reached the polling window with no terminal result → comment in Japanese that CI/review did not complete in the window, stop.

3. If the outcome is continue, wait the poll interval and start the next cycle automatically.

Loop parameters (ADR 0039)

• Poll interval: about 3 minutes between cycles.
• Polling window: at most 20 minutes total (about 6–7 cycles).
• Keep loop count and the last checked head SHA in session memory; track repair count with a todo per ADR 0039 §8.
• Never let the loop idle waiting for a human to manually re-trigger the next poll. The whole point of this skill is that "continue" reschedules itself.

Per-Agent Wait Mechanism

The only agent-specific part is how you wait the poll interval without blocking your runtime. Do not hardcode a single mechanism here; follow your agent guide:

• Claude Code → CLAUDE.md section "agent-pr-autoloop の Claude Code 実装" (ScheduleWakeup self-pacing; fallback bounded bash until loop; never sleep && <next>).
• Codex → AGENTS.md "Agent Git Workflow" (inline session polling; no ScheduleWakeup).

If your agent guide has no entry, fall back to the most conservative non-blocking poll your runtime supports and report which mechanism you used.

Stop Conditions

Stop the loop (and report the final monitored state) on merge, escalate, or timeout. Do not merge on continue, and do not treat an empty review/comment set as a clean review (it means "review not completed yet" per agent-pr-maintainer).

References

• .agents/skills/agent-pr-maintainer/SKILL.md — policy this skill executes.
• docs/decisions/0039-agent-pr-maintenance-automation.md — decision record.
• CLAUDE.md / AGENTS.md — per-agent wait mechanism.