pr-follow-up
Skill0 starsChanged 25 days agoMITView on GitHub
Follow up on PRs you authored: update the branch with its base and resolve conflicts, address review comments, and fix failed checks, then queue pushes, replies and thread resolution for confirmation. PRs given: those; one PR already discussed in the conversation: that one; "all" or nothing to go on: every open PR you authored across all your repos and PR hosts. Reuses resolve-conflicts, address-comments and qa. Use on "follow up on this PR", "follow up on my PRs", "babysit my PRs", "get my PRs green", "/pr-follow-up [urls]".
In this repository18 other files it gives its agents.
- address-commentsSkill
- demoSkill
- deploySkill
- documentSkill
- gather-contextSkill
- implementSkill
- investigateSkill
- observeSkill
- project-profileSkill
- pr-reviewSkill
- qa-planSkill
- qaSkill
- resolve-conflictsSkill
- specSkill
- sprintSkill
- standupSkill
- todoSkill
- triageSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
Follow ../../shared/policy.md.
- Very specific: line 45
Comments: run dev-skills:address-comments on the PR.
- Very specific: line 48
Failed checks: read the failing logs (how, per host: shared/references/pr-hosts.md).
- Very specific: line 51
Run dev-skills:qa in verify mode on everything changed in this PR.
- Distinctive: line 57
Stop and ask (unattended: put it in the report instead) when:
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
PR follow-up
Follow ../../shared/policy.md. Run dev-skills:project-profile first.
Host commands: ../../shared/references/pr-hosts.md.
1. Pick the PRs
- PR urls or numbers given: those.
- Ticket urls or keys given: each one's open PR (branch or title carries the key) on the project's PR host.
all, "my PRs", or a request about PRs in general: every open PR authored by me, in every repo, not just the current one:- GitHub, all repos:
gh search prs --author @me --state open --json repository,number,title,url, once per logged-in account (gh auth status). - Other hosts (Bitbucket, GitLab): each repo with a profile in
~/.dev-skills/profiles/, plus the current repo, one list call per repo.
- GitHub, all repos:
- Nothing given: the PR this conversation has been about (last PR url, number or branch discussed or worked on here). If none, ask once for the PR link, or whether to scan all your open PRs. Don't guess from the current branch.
Map each PR to its local checkout: the profile's Repo: path, else the current repo if it matches,
else a clone whose origin matches under the parent folders of known checkouts. None found: clone it
into ~/.dev-skills/repos/<project-slug> and run dev-skills:project-profile there.
With several PRs, batch the reads per repo: PR state, mergeability and checks in one call, unresolved
review threads in one query. Per PR, note what it needs: behind/conflicts, unaddressed
comments, failed checks. Skip PRs that need nothing. None left: say Nothing to follow up. and stop.
2. Per PR, in order
Load that repo's profile (dev-skills:project-profile from its checkout). Work in the PR's worktree
(reuse one under ~/.dev-skills/worktrees/, otherwise create it as in implement). Load the ticket
once via dev-skills:gather-context (reuse context.md).
- Update branch: merge
origin/<base>into the branch (never rebase a pushed branch). Conflicts go throughdev-skills:resolve-conflicts. - Comments: run
dev-skills:address-commentson the PR. Its rules apply: check each comment makes sense, replyDone! @<commenter>when fixed or a one-sentence reason when not, resolve bot threads and leave human ones open, and tag the ticket's author when a comment contradicts the ticket. - Failed checks: read the failing logs (how, per host:
shared/references/pr-hosts.md). Real failure: reproduce locally, fix, commit. Flaky or infra (timeout, runner, network, unrelated to the diff): queue a rerun instead of a code change. Failures caused by the base update or a comment change are fixed here too. - Run
dev-skills:qainverifymode on everything changed in this PR.
Fold every PR's push, replies, resolutions and reruns into one confirmation list for the whole run, grouped by PR. Push comes before the replies of the same PR.
Ask the user
Stop and ask (unattended: put it in the report instead) when:
- A fix, conflict resolution or comment would change behavior beyond what the comment or ticket says.
- A check keeps failing after one fix attempt, or its cause is unclear.
- Anything else blocks progress on a PR. Move on to the next PR while waiting when you can.
Report
One PR:
PR <repo>#<n> <title>: <up to date | merged base, N conflicts REVIEW>
Comments: <n done, n declined> · Checks: <ok | fixed X | rerun (flaky)>
Assumed: ...
<confirmation list>Several PRs:
<n> PRs followed up, <m> need you.
| Repo | PR | Branch | Comments | Checks | Notes |
|---|---|---|---|---|---|
| api | #12 <title> | merged base, 1 conflict REVIEW | 3 done, 1 declined | fixed lint | |
| web | #15 <title> | up to date | — | rerun (flaky) | asked: <question> |
Assumed: ...
<confirmation list>Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

