Skip to content
Skills Library
Programming · Review & audit

code-review

/code-review

Use when a named diff or pull request needs review for bugs, regressions, tests, and standards. Not for reviewing requirements or plans, handling feedback already left on a PR, or shipping the change.

Licence
Apache-2.0
Last repository change
10.11.2026

Assessment and security

90/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content25 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
2
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 10.11.2026, without references to retired models or APIs.

Licence

Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.

What it does and when to use it

Use when a named diff or pull request needs review for bugs, regressions, tests, and standards. Not for reviewing requirements or plans, handling feedback already left on a PR, or shipping the change.

Original author description

Use when a named diff or pull request needs review for bugs, regressions, tests, and standards. Not for reviewing requirements or plans, handling feedback already left on a PR, or shipping the change.

How it works

  1. Settle scope before reading code
  2. Choose lenses from risk
  3. Review against evidence
  4. Report verdict, findings, and limits
  5. Corrections only on explicit orders
  6. What you hand back
Original SKILL.md headings · 37 lines, 635 words
SKILL.md preview
---
name: code-review
description: 'Use when a named diff or pull request needs review for bugs, regressions, tests, and standards. Not for reviewing requirements or plans, handling feedback already left on a PR, or shipping the change.'
---

# Code review

A review hides no defect behind familiarity: every finding names its evidence. Judge the change against intended behavior and project requirements, not a preferred rewrite; a small diff can carry a serious defect, and an adequate change needs no findings. This skill is read only by default. It reports a verdict with findings and coverage, and it touches code only when the invocation carries an explicit apply authorization, only inside the reviewed scope.

No review here pushes, opens a PR, files a ticket, switches a branch, or checks out a working tree to reach its scope. Uncommitted work is reviewed only from the checkout that holds it.

## Settle scope before reading code

Pin down the reviewed scope first: an explicit revision range, a branch with a determinable base, or a PR identity. Resolve the base from explicit input or PR, upstream, or remote default evidence; never guess the base. Record what is in scope and what is explicitly out. Note the scope mode and depth signals too: full or focused, from the size, risk, and touched surfaces. Depth follows risk, never a fixed count. An unresolvable or unreadable scope means returning blocked with the missing identity or path, never dispatching a review at an unreachable target.

## Choose lenses from risk

Always cover correctness, regressions, tests, and the applicable standards for the touched surfaces. Add conditional lenses only on evidence, and name each selected lens with a one line reason:

*   Security: trust boundaries, data handling, or auth are touched.
*   Concurrency: shared state is touched.
*   Migration: schema, wire, or persisted shapes change.
*   Performance: only for measured hot paths, never on suspicion.
*   Control (agent or AI feature surfaces): the agent could act irreversibly. Check for an approval gate where one is needed, an escape hatch to interrupt or redirect, and visible confidence, escalation, and uncertainty signals; each finding cites the missing gate or signal with its consequence.

## Review against evidence

Every finding cites file and line, the offending code, the requirement or behavior it violates, and the consequence that justifies action. No invented issues and no style only findings. Check claims against the whole change and project facts; reviewer suspicion is not a finding until verified. Name the single fact the change's safety depends on; when no such fact exists, assess each risk independently. Trace consumers beyond symbol search, including library sources with pinned versions and local patches, scheduling and lifecycle ordering, and wire shapes, flags, and cross language readers up to three hops downstream, marking each path reachable with evidence or unreachable with the search that showed it. Record each risk with likelihood, cost, and exact file and line, and state plainly any caller or API not found in the tree. A lens with insufficient evidence is marked limited in coverage, not failed; nothing is invented to fill it.

## Report verdict, findings, and limits

Deliver the verdict, findings ordered by consequence, and explicit coverage: what was examined, which lenses ran, and the limits (unreadable paths, unavailable evidence, deferred surfaces). Note uncertainty rather than stopping to ask: this skill never blocks on questions. An adequate change gets the clean verdict with coverage, never manufactured findings.

## Corrections only on explicit orders

Report only is the default; a nested workflow later owns any repair. An explicit apply authorization in the invocation permits local corrections inside the reviewed scope with the narrowest applicable checks. Report the corrected files with their check results. An apply request without authority, or reaching beyond scope, is refused with the boundary stated.

## What you hand back

The verdict, findings with file and line evidence and consequences, coverage and limits, and, when apply ran, the corrected files with check results. No machinery narration follows, and no downstream action follows a report only review.
Files in this package (2)
  • SKILL.md 4 KB
  • agents/openai.yaml 1 KB

Versions

RepositoryScoreLicence
This page · gosuda/dalcom-skills90Apache-2.0
microsoft/aspire100MIT
kdlbs/kandev100AGPL-3.0
dotnet/runtime100MIT
drosam/agents100MIT
ggml-org/llama.cpp100MIT
ShamelesAbyss/Nyx100Apache-2.0
dzhng/jevgrep100MIT
iuliandita/skills100MIT

Database snapshot: 13.11.2026 · package version 0f70f0085480 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 90ce-code-review
    Review a named diff or PR for bugs, regressions, tests, and standards. Use when asked to review code or when a shipping skill needs a review receipt. Use when asked to apply this review's findings locally. Use ce-resolve-pr-feedback for feedback already left on a PR.
  • 90review
    Review a pull request or diff for this repository against otari's project standards. Use when asked to review a PR, inspect a diff, or prepare review feedback.
  • 80Code review
    Review a diff or module for correctness, security, regressions, and missing tests. Use when the user asks for a review or before shipping a change.
  • 100dev-review
    Use when a diff, branch or PR needs review before merge, on correctness, standards and the spec.
  • 90reviews
    Review a pull request or code change for code health. Use when a user asks to "code review", "review this PR", or "review this change".
  • 100review-pr
    Review a pull request for code quality and security issues. Use when the user wants to review a PR, check PR code changes, or audit a pull request. Triggers on: "review PR", "review pull request", "check this PR", "the `review-pr` skill".

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.