Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/skuridin-address-review
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/address-review
What's inside
- Skill name
address-review- Description
- Handle review comments on my own pull request. Checks out the PR, fetches unresolved review threads, verifies each claim against the code, recommends fix / reply / follow-up issue, then fixes and replies on GitHub once I decide. Use when asked to "fetch review comments and verify them", "address review", "respond to review comments", "check the review on my PR", or given a PR URL with a
- License
- Apache-2.0
- Instructions
- 52 lines in SKILL.md · view on GitHub
Files (2)
- SKILL.md
- pr-threads · scripts/
SKILL.md (first lines)
# Address Review
For PRs I authored. Reviewing someone else's PR is a different job (pr-review-bot / thermo-nuclear-pr-review).
## 1. Get on the branch
- Target: PR number or URL from the prompt, else the PR for the current branch.
- Not on it yet: `gh pr checkout <n>`. Dirty tree: stop and ask.
- Force-pushed remote: `git fetch` and `git reset --hard origin/<branch>` only after confirming no local commits would be lost.
## 2. Fetch threads
Run `scripts/pr-threads [n|url]` (relative to this skill's directory). Add `--all` to include resolved threads.
Each thread prints `thread <node-id> reply-to <comment-id>`, the file:line, and the conversation. Skip threads where I already replied and the reviewer has not answered.
## 3. Verify each claim
- Read the actual code at the cited location and trace the real flow. Do not trust the reviewer's description of the code.
- If the claim depends on another system (backend, API contract), say so and offer to check that codebase instead of guessing.
- More than two non-trivial claims: verify with an independent read-only subagent, give it the thread text and the files, ask for `REAL|NOT REAL|PARTLY` plus one line of evidence each.
Report per thread, short:
```
1. <file:line> <claim in plain words>
verdict: real / not real / partly — <evidence>
do: fix in this PR / reply only / follow-up issue — <why>
```
"Follow-up issue" when the claim is real but outside this PR's scope. Then stop and wait for my decision. Do not edit or post yet.
## 4. Act on my decision
- Fix: smallest change that addresses the root cause, then run the repo's verify command. Commit with the repo's commit convention.
- Push only when I say so.
- Reply only after the fix is pushed, so replies reference committed code (short SHA, file names).
- Follow-up issue: create it with `gh issue create`, What + Why sections, framed as something to investigate if no decision was made. Link it in the reply.
## 5. Reply on GitHub
```sh
gh api repos/{owner}/{repo}/pulls/<n>/comments/<reply-to>/replies -f body='...'
```
- 1-3 sentences, casual peer tone, lowercase fine, plain words. Must not read as AI-written: no "great catch!", no bullet lists, no headings.
- Say what was done and where (`fixed in abc1234, ...`), or why it stays as is with the evidence.
- No bot footer: these go out in my voice, so draft them and show me before posting unless I already said "reply".
- Do not resolve threads; the reviewer does that.
- A POST can error after it succeeded. Re-run `pr-threads` to check before posting again.From skuridin/agent-stuff, Apache-2.0.
More from skuridin/agent-stuff
- 0.561score
publish-html-report
skuridin
Use this skill whenever you create or finish a standalone HTML report or presentation intended for the requester to view in a browser.
SkillAFreshness: AClaude Code +9 - 0.529score
pr-review-bot
skuridin
Unattended thermo-nuclear PR review loop.
SkillBFreshness: BClaude Code +9 - 0.526score
thermo-nuclear-pr-review
skuridin
Thermo-nuclear code quality review of a GitHub pull request.
SkillBFreshness: BClaude Code +9 - 0.474score
panel-review
skuridin
Use this skill when reviewing code, PRs, files, or directories as an expert panel; supports interactive walkthroughs and opt-in draft GitHub reviews for…
SkillCFreshness: CClaude Code +9 - 0.474score
bearcli
skuridin
Use Bear CLI to read, search, create, and safely edit Bear notes from the terminal.
SkillCFreshness: CClaude Code +9 - 0.469score
frontend-instrumented-debugging
skuridin
Use this skill when debugging hard JavaScript frontend bugs where runtime behavior is unclear, especially async flows, state gates, retries, effects…
SkillCFreshness: CClaude Code +9 - 0.455score
apple-container
skuridin
Use this skill when the user wants to work with Apple's `container` CLI for macOS — building, running, or managing lightweight Linux containers using…
SkillCFreshness: CClaude Code +9 - 0.445score
grill-me
skuridin
Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree.
SkillCFreshness: CClaude Code +9 - 0.409score
functional-spec
skuridin
Write durable Functional Specifications that describe desired behavior, constraints, and acceptance criteria without coupling the document to implementation…
SkillDFreshness: DClaude Code +9 - 0.409score
vertical-slice-tracker
skuridin
Break implementation work into end-to-end vertical slices, create or update a markdown task tracker when the user asks for written planning, and keep each…
SkillDFreshness: DClaude Code +9 - 0.409score
gondolin-exec
skuridin
Run commands in an isolated, disposable Linux VM using Gondolin CLI.
SkillDFreshness: DClaude Code +9 - 0.409score
pr-create
skuridin
Create GitHub pull requests with conventional commit titles and clear, non-technical descriptions answering what changed and why.
SkillDFreshness: DClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Sep 27, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/skuridin/agent-stuff
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from skuridin
- 0.561score
publish-html-report
skuridin
Use this skill whenever you create or finish a standalone HTML report or presentation intended for the requester to view in a browser.
SkillAFreshness: AClaude Code +9 - 0.529score
pr-review-bot
skuridin
Unattended thermo-nuclear PR review loop.
SkillBFreshness: BClaude Code +9 - 0.526score
thermo-nuclear-pr-review
skuridin
Thermo-nuclear code quality review of a GitHub pull request.
SkillBFreshness: BClaude Code +9 - 0.474score
panel-review
skuridin
Use this skill when reviewing code, PRs, files, or directories as an expert panel; supports interactive walkthroughs and opt-in draft GitHub reviews for…
SkillCFreshness: CClaude Code +9 - 0.474score
bearcli
skuridin
Use Bear CLI to read, search, create, and safely edit Bear notes from the terminal.
SkillCFreshness: CClaude Code +9 - 0.469score
frontend-instrumented-debugging
skuridin
Use this skill when debugging hard JavaScript frontend bugs where runtime behavior is unclear, especially async flows, state gates, retries, effects…
SkillCFreshness: CClaude Code +9
Cite this page
address-review skill. Agent Mart. https://agentmart.dev/skills/skuridin-address-review (last updated 2026-10-24)