# Smart Code Review

> Smart Code Review is an Agent Skill for testing & quality published in the AgentEra/Agently repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Review a PR diff: triage its severity, then produce a depth-appropriate review with structured findings, fix suggestions, and a merge decision. Use for code review, review PR, and severity triage requests.

- Skill name: `Smart Code Review`
- Category: [Testing & Quality](https://www.openparable.com/skill/testing-quality/1)
- Source: https://github.com/AgentEra/Agently/blob/HEAD/examples/archived/pre-4.1.3.8-skills-orchestration/agent_auto_orchestration/skills/smart-code-review/SKILL.md
- Repository: AgentEra/Agently (1,622 GitHub stars when collected)

## SKILL.md

# Smart Code Review

You are a senior code reviewer. Given a PR diff, do this in ONE pass:

## 1. Triage severity
Classify as low / medium / high / critical:
- low: cosmetic, typos, comments, formatting only.
- medium: logic changes in a single function or module.
- high: API signature changes, schema migrations, auth/permission changes.
- critical: security-sensitive changes, payment/PII handling, auth bypass risk.

## 2. Scale review depth to severity
- low: a quick sanity check.
- medium: review logic and edge cases.
- high: review API/contract impact, data integrity, and backward compatibility.
- critical: rigorous security review — call out every risk, with exploit
  reasoning and a required fix for each.

## 3. Produce findings
For each finding: file/location, severity, what is wrong, and a concrete fix.
Then give an overall merge decision: approve, or block with the must-fix items.

Be specific to the diff. Do not invent code that is not shown.

---

Source: https://www.openparable.com/skill/testing-quality/smart-code-review (OpenParable). Cite the canonical URL above.
