code-review

Name: code-review
Author: selmakcby

111stars17forks

How to install code-review?

selmakcby/claude-agents-skills/code-review

$npx -y skills add selmakcby/claude-agents-skills --skill code-review

Installs into the current project.

Summary

Senior code review specialist. Use PROACTIVELY after writing or modifying code. Reviews for quality, correctness, performance, and best practices. Assigns severity levels to findings (CRITICAL / HIGH / MEDIUM / LOW).

Files · 1

View on GitHub

SKILL.md

1	<!--
2	Source: anthropics/claude-code (official)
3	File: https://github.com/anthropics/claude-code/blob/main/plugins/code-review/commands/code-review.md
4	Also: https://github.com/anthropics/claude-code/blob/main/plugins/code-review/README.md
5	Used by: reviewer agent
6	-->
7
8	# Code Review
9
10	## When to trigger
11	- Immediately after new or modified code
12	- Before commit / PR
13	- When user asks for `/code-review`
14
15	## Review approach
16
17	Multi-agent review with confidence filtering:
18
19	1. CLAUDE.md compliance — does the code follow project rules?
20	2. Bug detector — obvious bugs in the diff
21	3. History analyzer — git context to understand intent
22
23	Each issue is scored 0–100. Findings below threshold (default 80) are filtered out.
24
25	## What to check
26
27	### Clarity
28	- Names (variables, functions, files) — do they say what they do?
29	- Structure — is it easy to scan?
30	- Cognitive load — how much do you hold in your head to understand it?
31
32	### Correctness
33	- Edge cases: null, undefined, empty array, boundary values
34	- Error paths: are they handled? Do they fail loudly or silently?
35	- Async: await missing? Race conditions? Promise rejections caught?
36	- Off-by-one errors in loops / slicing
37
38	### Performance
39	- N+1 queries (especially around ORM usage)
40	- O(n²) algorithms on large collections
41	- Unnecessary re-renders (React)
42	- Missing memoization for expensive computations
43	- Synchronous I/O in hot paths
44
45	### Consistency
46	- Matches project style guide (CLAUDE.md, rules/)
47	- Immutable patterns used
48	- Function size under 50 lines
49	- Nesting 4 levels max
50
51	### Tests
52	- Is there a test for the new behavior?
53	- Edge cases covered?
54	- Mocks realistic (no "happy path only")?
55	- Tests independent (no shared state)?
56
57	### Dead code
58	- Unused imports
59	- Unreachable branches
60	- TODO comments that should be done or deleted
61	- Commented-out code
62
63	## Severity levels
64
65	- CRITICAL — do not commit. Breaks production, security hole, data loss risk.
66	- HIGH — fix before next release. Real bug users will hit.
67	- MEDIUM — fix when you're in the area. Code smell, minor inefficiency.
68	- LOW — nice-to-have. Style preference, optional refactor.
69
70	## Output format
71
72	```markdown
73	## CRITICAL (must-fix)
74	- [`<file>:<line>`] <issue>
75	- Why: <impact>
76	- Fix: <specific suggestion>
77
78	## HIGH
79	- ...
80
81	## MEDIUM / LOW
82	- ...
83
84	## Summary
85	<N CRITICAL · N HIGH · N MEDIUM · N LOW>
86	<Verdict: PASS \| FAIL>
87	```
88
89	## Rules
90
91	- Every finding has: severity, file + line, what's wrong, why it matters, suggested fix.
92	- Don't just flag — propose a fix.
93	- Don't be timid on CRITICAL. If it's critical, say so clearly.
94	- If you see a pattern (same issue in multiple places), flag it once with a list.
95	- Use `--comment` mode for PR review, default for local output.

People who liked this also viewed

6 picks

Type

code-review

111stars17forks

View on GitHub

How to install code-review?

selmakcby/claude-agents-skills/code-review

$npx -y skills add selmakcby/claude-agents-skills --skill code-review

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	<!--
2	Source: anthropics/claude-code (official)
3	File: https://github.com/anthropics/claude-code/blob/main/plugins/code-review/commands/code-review.md
4	Also: https://github.com/anthropics/claude-code/blob/main/plugins/code-review/README.md
5	Used by: reviewer agent
6	-->
7
8	# Code Review
9
10	## When to trigger
11	- Immediately after new or modified code
12	- Before commit / PR
13	- When user asks for `/code-review`
14
15	## Review approach
16
17	Multi-agent review with confidence filtering:
18
19	1. CLAUDE.md compliance — does the code follow project rules?
20	2. Bug detector — obvious bugs in the diff
21	3. History analyzer — git context to understand intent
22
23	Each issue is scored 0–100. Findings below threshold (default 80) are filtered out.
24
25	## What to check
26
27	### Clarity
28	- Names (variables, functions, files) — do they say what they do?
29	- Structure — is it easy to scan?
30	- Cognitive load — how much do you hold in your head to understand it?
31
32	### Correctness
33	- Edge cases: null, undefined, empty array, boundary values
34	- Error paths: are they handled? Do they fail loudly or silently?
35	- Async: await missing? Race conditions? Promise rejections caught?
36	- Off-by-one errors in loops / slicing
37
38	### Performance
39	- N+1 queries (especially around ORM usage)
40	- O(n²) algorithms on large collections
41	- Unnecessary re-renders (React)
42	- Missing memoization for expensive computations
43	- Synchronous I/O in hot paths
44
45	### Consistency
46	- Matches project style guide (CLAUDE.md, rules/)
47	- Immutable patterns used
48	- Function size under 50 lines
49	- Nesting 4 levels max
50
51	### Tests
52	- Is there a test for the new behavior?
53	- Edge cases covered?
54	- Mocks realistic (no "happy path only")?
55	- Tests independent (no shared state)?
56
57	### Dead code
58	- Unused imports
59	- Unreachable branches
60	- TODO comments that should be done or deleted
61	- Commented-out code
62
63	## Severity levels
64
65	- CRITICAL — do not commit. Breaks production, security hole, data loss risk.
66	- HIGH — fix before next release. Real bug users will hit.
67	- MEDIUM — fix when you're in the area. Code smell, minor inefficiency.
68	- LOW — nice-to-have. Style preference, optional refactor.
69
70	## Output format
71
72	```markdown
73	## CRITICAL (must-fix)
74	- [`<file>:<line>`] <issue>
75	- Why: <impact>
76	- Fix: <specific suggestion>
77
78	## HIGH
79	- ...
80
81	## MEDIUM / LOW
82	- ...
83
84	## Summary
85	<N CRITICAL · N HIGH · N MEDIUM · N LOW>
86	<Verdict: PASS \| FAIL>
87	```
88
89	## Rules
90
91	- Every finding has: severity, file + line, what's wrong, why it matters, suggested fix.
92	- Don't just flag — propose a fix.
93	- Don't be timid on CRITICAL. If it's critical, say so clearly.
94	- If you see a pattern (same issue in multiple places), flag it once with a list.
95	- Use `--comment` mode for PR review, default for local output.

People who liked this also viewed

6 picks

Type