issue-review — Claude Code skill: label, analyze, deduplicate issues
A skill for Claude Code, Codex, Cursor and other agents that fetches open GitHub Issues, reads related source code, assigns labels, and groups duplicates. Trigger it with /issue-review from your agent chat.
MIT 208 installs ★ 11,397 updated 2026-10-24
npx skills add toss/es-toolkit --skill issue-review -a claude-codeWhat issue-review does
The skill calls gh issue list to pull open issues up to the specified count, then reads each issue's body and comments for a per-issue deep review.
For every issue that mentions a function, the agent opens the actual source file and existing tests to judge whether the reported behavior is a bug or by design — including a lodash comparison for bug reports. Feature requests are evaluated against project design principles and TC39 stage. Issues without labels get /issue-label {number} applied automatically. Duplicates surface through gh issue list --search by function name and similar error descriptions.
The command defaults to 10 issues; pass a number as the argument for a larger batch. GitHub CLI (gh) must be installed and authenticated against the target repository.
What it helps with
- Label a batch of new issues without manual triage
- Check whether a reported behavior is a bug or intentional design
- Find duplicate issues filed against the same function
- Evaluate a feature request against the project's design principles
- Get a summary table of reviewed issues with recommended actions
Example prompts for the agent
- Review the 10 most recent open issues and apply labels:
/issue-review - Triage 20 issues and surface duplicates:
/issue-review 20 - Run a full issue review and flag candidates for closing as wontfix:
/issue-review
How to install issue-review
1Skills CLI (Vercel)
npx skills add toss/es-toolkit --skill issue-review -a claude-code2Manually: put the files in place
~/.claude/skills/issue-review/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/issue-review/
1Skills CLI (Vercel)
npx skills add toss/es-toolkit --skill issue-review -a codex2Manually: put the files in place
.agents/skills/issue-review/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add toss/es-toolkit --skill issue-review -a cursor2Manually: put the files in place
.cursor/skills/issue-review/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install toss/es-toolkit .claude/skills/issue-review2Manually: put the files in place
.github/skills/issue-review/SKILL.md1Skills CLI (Vercel)
npx skills add toss/es-toolkit --skill issue-review -a gemini-cli2Platform CLI
gemini skills install https://github.com/toss/es-toolkit --path .claude/skills/issue-reviewManually: put the files in place
.devin/skills/issue-review/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add toss/es-toolkit --skill issue-review -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about issue-review
How do I install issue-review in Claude Code?
Run npx skills add toss/es-toolkit --skill issue-review -a claude-code. The skill file will be placed at ~/.claude/skills/issue-review/SKILL.md.
What does the issue-review skill do?
It fetches open issues via GitHub CLI, reads the source code of any mentioned functions, classifies each issue by type, assigns a label if none exists, and detects duplicates across the repository.
What does issue-review need to work?
GitHub CLI (gh) must be installed and authenticated for the target repository. The skill uses Bash, Read, Grep, Glob, and Skill — all of which must be available to the agent.
Source text
SKILL.md · MIT
---
name: issue-review
description: Review recent issues with labeling, context analysis, and duplicate detection
argument-hint: '[count=10]'
allowed-tools: Bash, Read, Grep, Glob, Skill
---
# Issue Review
Fetch recent issues, label unlabeled ones with context, detect duplicates.
## Input
$ARGUMENTS — Number of issues (default: `10`)
Examples:
- `/issue-review` — 10 most recent open issues
- `/issue-review 20` — 20 issues
## Workflow
### 1. Fetch recent issues
```bash
gh issue list --repo toss/es-toolkit --state open --limit {count} --json number,title,author,labels,createdAt
```
### 2. Deep review per issue
For each issue:
#### a. Read issue content
```bash
gh issue view {number} --repo toss/es-toolkit --json title,body,labels,comments
```
#### b. Read related source code
If the issue mentions a specific function:
- Read the function source to understand current behavior
- Read existing tests to see what's covered
- Check if there's already a compat variant
#### c. Provide context
- **Bug reports**: Is the reported behavior actually a bug? Or is it by design? Does lodash behave differently?
- **Feature requests**: Does this align with design principles? Is it replaceable by modern JS? Is it TC39 Stage 3+?
- **Type issues**: Read the current type signature, assess the proposed change
- **Docs**: Check what's currently documented vs what's being requested
#### d. Label if unlabeled
If no labels exist, run `/issue-label {number}`.
### 3. Detect duplicates
```bash
gh issue list --repo toss/es-toolkit --state all --search "{function name}" --limit 10 --json number,title,state,labels
```
Group by:
- Same function name in title
- Similar error descriptions
- Same feature being requested
### 4. Report per issue
```
### Issue #{number} — {title}
**Label**: {existing or newly applied}
**Context**: {what the function currently does, relevant code snippet}
**Analysis**: {is the request valid? design principle alignment?}
**Duplicates**: {similar issues if any}
**Action**: {label applied / needs discussion / close as wontfix / link to existing PR}
```
### 5. Summary
```
## Issue Review — {date}
| # | Title | Label | Duplicate? | Action |
|---|-------|-------|------------|--------|
- {N} issues reviewed
- {N} newly labeled
- {N} potential duplicates
- {N} actionable bugs
- {N} feature requests
```
© toss · MIT · toss/es-toolkit