address-comments
Skill0 starsChanged 25 days agoMITView on GitHub
Work through code review comments on your own PR: check each one makes sense against the requirements and code, implement the valid ones, verify, commit, then queue push, short replies and thread resolution (bots resolved, humans left open) for confirmation. Any PR host. Use on "address the review comments", "address CR comments", "fix PR feedback", "respond to review", "/address-comments".
3 other projects use a version of this.
Adopted by
Most recently taken up by
- bryce-evans/claude-skillstheir own version, first seen Oct 13, 2026
- imbue-ai/sculptortheir own version, first seen Oct 8, 2026
- oleg-chibikov/agent-skillstheir own version, first seen Sep 16, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- imbue-ai/sculptor★ 246same name
- oleg-chibikov/agent-skills★ 1same name
- bryce-evans/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository18 other files it gives its agents.
- demoSkill
- deploySkill
- documentSkill
- gather-contextSkill
- implementSkill
- investigateSkill
- observeSkill
- pr-follow-upSkill
- 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
4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Follow ../../shared/policy.md.
- Distinctive: line 32
Contradicts the ticket: no change yet; reply tagging the ticket's author to confirm (their PR host handle if known, otherwise also comment …
- Very specific: line 36
Ask the user (unattended: list it) when following a comment would change behavior well beyond what the comment or ticket says, or when …
- Very specific: line 44
If the branch needs the latest base to pass, merge it in (conflicts via dev-skills:resolve-conflicts).
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Address comments
Follow ../../shared/policy.md. Run dev-skills:project-profile first.
Host commands: ../../shared/references/pr-hosts.md.
1. Fetch
- The PR: given, or the open PR for the current branch.
- Unresolved review threads and general comments not yet answered by you. Skip resolved, deleted and
outdated threads whose code no longer exists (mention them). None left: say
Nothing to address on <PR>.and stop, before any worktree or ticket work. - Work in its branch (its existing worktree under
~/.dev-skills/worktrees/if there is one; otherwise create one as inimplement). - Requirements: the ticket via
dev-skills:gather-context(reusecontext.mdif present).
2. Decide per comment
Tell humans from bots: GitHub author.__typename == "Bot" or a login ending in [bot]; known review
bots (CodeRabbit, Copilot, Sonar, Codecov...) count as bots. Then check whether the comment makes
sense against the code and the requirements:
- Valid: implement it, including optional suggestions and nits.
- Not valid (wrong about the code, would introduce a bug, or adds a large change the ticket does not need): no change; reply with the reason in one short sentence.
- Contradicts the ticket: no change yet; reply tagging the ticket's author to confirm (their PR host handle if known, otherwise also comment on the ticket mentioning them). Keep it open.
- Question: answer it; change code only if the answer reveals a problem.
- Two comments conflicting with each other: follow the more senior/blocking reviewer's and note it.
- Ask the user (unattended: list it) when following a comment would change behavior well beyond what
the comment or ticket says, or when stuck.
Declines follow the mode:
autonomousdecide and record;balanceddecide and show in the confirmation list;guidedask per comment.
3. Change, verify, commit
Implement the accepted changes. Run dev-skills:qa in verify mode on what changed. Commit locally
per the mode, one commit per logical change or one commit for all, following the repo's history style.
If the branch needs the latest base to pass, merge it in (conflicts via dev-skills:resolve-conflicts);
rebasing a pushed branch is a confirmation-list item.
4. Queue
Replies are short, especially to humans. Per thread:
| Comment | Reply | Thread |
| Valid, done | Done! @<commenter> (nothing else) | human: leave open · bot: resolve |
| Not valid | @<commenter> <reason, one sentence> | human: leave open · bot: resolve |
| Contradicts ticket | @<commenter> this goes against <TICKET> (<what>). @<ticket author> can you confirm? | leave open |
- Push first; replies go after the push.
- General (non-thread) comments: reply with a PR comment; there is nothing to resolve.
- Re-request review from humans if the host supports it.
Report
PR <link>
| # | Comment (short) | Decision | Change |
|---|---|---|---|
| 1 | `file:line` "use a transaction here" | done | <sha> |
| 2 | "rename X" | declined | <reason> |
Checks: <command> → pass/fail
<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.

