Skip to content
Skills Library
Programming · Review & audit

code-review

/code-review

Review criteria and confidence scoring for changes in the harness-eval monorepo (Bash scripts, hooks, SKILL/agent prompts, manifests). Use when reviewing a diff, PR, or file set in this repo.

Licence
MIT
Last repository change
24.09.2026

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
1
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 24.09.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Review criteria and confidence scoring for changes in the harness-eval monorepo (Bash scripts, hooks, SKILL/agent prompts, manifests). Use when reviewing a diff, PR, or file set in this repo.

Original author description

Review criteria and confidence scoring for changes in the harness-eval monorepo (Bash scripts, hooks, SKILL/agent prompts, manifests). Use when reviewing a diff, PR, or file set in this repo.

How it works

  1. Review Scope
  2. What to check in this repo
  3. Confidence and severity
  4. Output Format
Original SKILL.md headings · 35 lines, 374 words
SKILL.md preview
---
name: code-review
description: Review criteria and confidence scoring for changes in the harness-eval monorepo (Bash scripts, hooks, SKILL/agent prompts, manifests). Use when reviewing a diff, PR, or file set in this repo.
---

# Code Review Skill

Review changed code in this repo and report every issue with a severity and a confidence score.

## Review Scope

By default, review all uncommitted work: `git status --short --untracked-files=all` for the file list, `git diff HEAD` for tracked changes (staged and unstaged), and each untracked (`??`) file read in full, since a new file has no diff. The user may specify different files or scope.

## What to check in this repo

Beyond general correctness and security, the contracts most likely to break here (plugin paths are under `plugins/harness-eval/`):
- Scripts: target project root as `$1` (`aggregate.sh` reads stdin instead), JSON on stdout / logs on stderr, exit codes 0/1/2, `jq` checked at startup (`scripts/CLAUDE.md`). The SKILL.md files duplicate this contract for plugin users, so a script change must update them.
- Prompt files are code. SKILL, agent, and command `.md` files carry contracts other code parses - handoff lines and markers, report headings, JSON field names, headings asserted by `tests/structure/test-plugin-structure.sh`. A changed string must match every consumer; grep for it.
- Auto-discovery: no non-component `.md` (including CLAUDE.md) under `commands/` or `agents/`. Agents use `tools`; commands and skills use `allowed-tools`.
- The version must match in `plugin.json` `version` and `marketplace.json` `metadata.version` + `plugins[].version`.
- Plugin hooks must not modify user files without opt-in, and their failures stay visible (`hooks/CLAUDE.md`).
- The fixtures under `tests/fixtures/` are frozen test inputs; new coverage goes in a new fixture directory.
- Bilingual output: English and Korean reports and doc sections stay in sync.

## Confidence and severity

For each issue give a severity (critical / important / minor) and a confidence 0-100 that it is a real, newly introduced defect (90+: confirmed; 75-89: verified; 50-74: likely; below 50: speculative or pre-existing). Report every issue with its scores. The report shows items at 75+ first and groups the rest under "Lower confidence" - the reader filters, you don't drop.

## Output Format

For each issue:
### [CRITICAL|IMPORTANT|MINOR] <issue title> (confidence: XX)
**File:** `path/to/file.ext:line`
**Issue:** Clear description of the problem
**Guideline:** Reference to CLAUDE.md rule or security standard
**Fix:** Concrete code suggestion

List items with confidence ≥ 75 first, then a "Lower confidence" section; if nothing reaches 75, say so and still list the rest.

Versions

RepositoryScoreLicence
This page · whchoi98/harness-eval90MIT
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 6ed0e8bebe8d (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

  • 92write-script-bash
    MUST use when writing Bash scripts.
  • 90review-code
    Code review flow — review a diff, PR, or code change, delegating the review to a subagent that follows the code-review-criteria skill. Use it when the user asks to review code or a PR, in any phrasing.
  • 90refactor
    Restructure the aws-blog-skills skill files (SKILL.md, references/, scripts/) or the bash harness without changing behavior, verifying with tests/run-all.sh after each step. Use when asked to reorganize, deduplicate, shorten, or split the skill, its scripts, or the harness.
  • 90bash-script
    Use when writing or editing a Bash script (a `.sh` file or one with a Bash shebang). Covers the template, conventions, and required patterns.
  • 100plugin-review
    Review changes to the aws-blog-skills plugin (skills/aws-blog-skills SKILL.md, references, scripts, .claude-plugin manifests, README) with confidence-based scoring, reporting only issues >= 75. Use when asked to review a diff or changes before commit or release. Named…
  • 100review-shell-scripts
    Review all shell scripts against the Bash Script Convention. Use when you need to check script compliance or before committing changes.

Feedback and reports

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