pull-request-follow-up

name	pull-request:follow-up
description	Follow up as the author on review feedback on your PR/MR: triage and draft replies, catch silent resolves, and drive bot reviewers to a passing score. Triggers: "address reviewer feedback", "make the bot reviewer pass", "greptile/coderabbit review", "bot review before pushing".
argument-hint	[pr-url] [--auto] [--include-human-nits] [--local [base]]
allowed-tools	["Bash(git:*)","Bash(gh:*)","Bash(greptile:*)","Bash(coderabbit:*)","Bash(jq:*)","Bash(bun ${CLAUDE_PLUGIN_ROOT}/scripts/*)","Skill(github:pr-comments)","Skill(gitlab:merge-request)","Skill(pull-request:babysit)","mcp__github"]

PR Follow-Up

Follow up on review feedback for: $ARGUMENTS

Parse the URL and flags from $ARGUMENTS. GitHub is primary: work through gh, mcp__github, and github:* skills directly. Delegate all GitLab behavior to gitlab:merge-request (no glab calls).

Arguments

• --auto: autonomously triage bot threads, looping until the acceptance bar is met (see The Autonomous Loop). Human threads stay gated.
• --include-human-nits: under --auto, also act on human threads, but only trivial high-confidence changes (typos, renames, one-liners). Off by default.
• --local [base]: run the bot loop pre-push through the reviewer's CLI instead of PR threads, against the branch's unmerged commits (see Local Mode). The optional base overrides what the review runs against. No PR is involved, so pr-url and the other flags don't apply.

With no flags, run the gated default.

Default Workflow (Gated)

Without --auto, stay read-only: fetch, classify, draft, check with me before posting or resolving.

Fetch all resolvable threads (resolved and unresolved) via github:pr-comments (--role author) or gitlab:merge-request, then classify:

• Unresolved / no reply
• Unresolved / replied
• Resolved / with reply
• Resolved / silent: resolved with no author reply; flag these, they hide context

For unresolved threads, diff the comment's creation date against HEAD to see whether later commits addressed the feedback. Draft replies per replies.md.

The Autonomous Loop

With --auto, drive bot threads to closure without asking. Partition threads with each provider's --bots filter (github:pr-comments, gitlab:merge-request): fetch once filtered for the bot set, once unfiltered to see the human threads you're leaving alone. Add reviewers per reviewers.md.

Triage each bot thread:

• Actionable → fix in the working tree, batching across threads. For a straightforward fix (test coverage, a rename, an obvious guard), acknowledge with a thumbs up reaction rather than a fleshed-out reply (see github:pr-comments, Reactions)
• Noise / false positive → reply with a one-line reason and resolve, or thumbs down a clearly wrong bot comment
• Unsure → collect to escalate; don't guess

Each round: apply the batched fixes and push once, reply-and-resolve the noise, escalate the unsure threads and pause that subset only, then get CI green and the bot onto the green SHA. auto.md has the round mechanics: the babysit handoff, the wake cadence while a re-review lands, and repos that review only on request.

Acceptance Bar

The loop ends when the reviewer's own score on the current HEAD is at its maximum (Greptile 5/5, CodeRabbit Actionable comments posted: 0), or when every comment still standing below that maximum carries a written reply giving the reason it was declined. Read the score off the summary comment on HEAD (reviewers.md): a score from an earlier SHA is stale, and an absent summary where a review is expected means the review is still running, so keep waiting for it.

A score short of the maximum with a silently skipped comment behind it does not clear the bar. Fix the comment or write the reason.

Guards stop a loop that can't converge: no new bot threads for two rounds after a green push, four rounds total, an idle timeout that outlasts the re-trigger, or the PR closing or merging. A guard ends the run without clearing the bar, so report the score reached and each comment left standing without a reason.

On stop, report fixes, replies/resolves, and escalations. Once the bar is clear, suggest the next action (human review, merge train, auto-merge) but don't perform it unless asked. pull-request:babysit --merge drives to merged.

Local Mode (Pre-Push)

With --local, run the same reviewer against the branch's unmerged commits before anything is pushed. Criteria and acceptance bar don't change with the channel: findings arrive as CLI output instead of PR threads, and the score in its local form (local.md) is still the exit. A declined finding goes in the report with its reason instead of a resolved thread.

Fire this mode unprompted when you are about to push or open a PR in a repo with an available bot and the diff warrants a metered review (local.md sets that gate). /ship runs this as a gated pass and pull-request:create runs it before pushing, so skip it when either already ran on this branch.

• Local detection: !bun ${CLAUDE_PLUGIN_ROOT}/scripts/detect-bot.ts

The line above is the injected fast path (repo config, CLI presence, and any live cooldown, no turn spent). Resolve it to a provider per local.md, which also covers the CLI mechanics and the hosted signals for repos with no config file. A provider reported as paused is out: report the pause and stop.

Then loop: run the review per local.md, summarize findings by severity with a file:line reference each, triage with the same partition as --auto, commit the fixes, and re-run until the bar is clear or I stop. Surface a disagreement with your reasoning rather than skipping it silently. Finish by offering next steps (push, PR, /ship) without taking them.

Once a PR exists, triage the hosted bot's comments through the flow above.

Guardrails

The gated default requires checking before any post or resolve. --auto lifts this for bot threads only; human threads stay gated unless --include-human-nits. Always:

• Resolve only after replying or reacting; silent resolves hide context.
• Never name or thank the bot in a reply; write it as a note for any reader (replies.md). The @<bot> re-trigger is the one exception.
• Thumbs down is bot-only feedback. Never thumbs down a human thread, autonomously or otherwise.
• Match my writing style; you're replying as me.