Skip to content
Skills Library
Programming · Review & audit

code-review-iterative

/code-review-iterative

Run iterative code review with diminishing fix thresholds. Use when the user says "review my code", "iterative review", "code review", "review this iteratively", or after completing a feature implementation.

Repository
mstack ↗
Licence
MIT
Last repository change
12.10.2026

Assessment and security

100/ 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.
Content35 / 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 12.10.2026, 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

Run iterative code review with diminishing fix thresholds. Use when the user says "review my code", "iterative review", "code review", "review this iteratively", or after completing a feature implementation.

Triggers from the description

review my codeiterative reviewcode reviewreview this iteratively
Original author description

Run iterative code review with diminishing fix thresholds. Use when the user says "review my code", "iterative review", "code review", "review this iteratively", or after completing a feature implementation.

How it works

  1. Review Scope
  2. Process
Original SKILL.md headings · 93 lines, 769 words
SKILL.md preview
---
name: code-review-iterative
description: Run iterative code review with diminishing fix thresholds. Use when the user says "review my code", "iterative review", "code review", "review this iteratively", or after completing a feature implementation.
---

# Iterative Code Review

Run up to 4 review-fix iterations with 3 parallel reviewer agents (PE, Sr SDE, QA). Each iteration raises the bar for what gets fixed, converging quickly to clean code.

## Review Scope

By default, review unstaged changes from `git diff`. If the user specifies files, a branch, or a PR, use that scope instead.

## Process

### Step 0: Initialize

Set `iteration = 1`. Determine the review scope (git diff, specific files, or user-specified).

### Step 1: Launch 3 Reviewer Agents in Parallel

Dispatch 3 agents simultaneously, each reviewing the same code from a different perspective. Every agent MUST categorize each finding as exactly one of: **BLOCKER**, **HIGH**, **MEDIUM**, **LOW**, **NIT**.

**Agent 1 — Principal Engineer (PE):**
> Review this code as a Principal Engineer. Focus on: architecture, design patterns, abstraction quality, API contracts, performance implications, scalability concerns, and whether the approach is fundamentally sound. Categorize every finding as BLOCKER, HIGH, MEDIUM, LOW, or NIT. Include file path and line number for each.
>
> **Stay in your lane:** You own architecture, AST/IR design, parser/compiler structure, and performance. Do NOT rule on whether a domain concept is semantically valid or whether a condition should accept certain inputs from a domain perspective — that is the Domain Expert's call. If you think a domain assumption in the code is wrong, flag it as a question rather than a finding.

**Agent 2 — Senior SDE:**
> Review this code as a Senior Software Engineer. Focus on: correctness, edge cases, error handling, null/undefined safety, race conditions, resource leaks, naming, readability, DRY violations, and adherence to project conventions (check CLAUDE.md). Categorize every finding as BLOCKER, HIGH, MEDIUM, LOW, or NIT. Include file path and line number for each.
>
> **Stay in your lane:** You own implementation correctness, code quality, and codebase consistency. Do NOT rule on domain semantics or override domain-driven design decisions. If a pattern seems wrong but is documented as a domain requirement, flag it as a question rather than a finding.

**Agent 3 — QA Engineer:**
> Review this code as a QA Engineer. Focus on: test coverage gaps, untested edge cases, missing error path tests, input validation, integration risks, regression potential, and observable behavior correctness. Categorize every finding as BLOCKER, HIGH, MEDIUM, LOW, or NIT. Include file path and line number for each.
>
> **Stay in your lane:** You own test coverage, test quality, and behavioral correctness. Do NOT rule on architectural decisions or domain semantics. If you see missing test cases that depend on domain knowledge (e.g., "should this condition accept lord references?"), flag it as a question for the Domain Expert rather than asserting the answer.

### Step 2: Consolidate and Filter

Merge findings from all 3 agents. Deduplicate (same issue found by multiple reviewers = single finding at highest severity). Apply the fix threshold for the current iteration:

| Iteration | Fix Threshold | What Gets Fixed |
|-----------|--------------|-----------------|
| 1 | BLOCKER, HIGH, MEDIUM, LOW, NIT | Everything |
| 2 | BLOCKER, HIGH, MEDIUM | Skip LOW, NIT |
| 3 | BLOCKER, HIGH | Skip MEDIUM and below |
| 4 | BLOCKER only | Skip HIGH and below |

Present the consolidated findings to the user in a table:

```
## Review Iteration {N} — Findings

| # | Severity | Reviewer | File:Line | Finding | Fix? |
|---|----------|----------|-----------|---------|------|
| 1 | BLOCKER  | PE       | src/x.rs:45 | ... | Yes |
| 2 | HIGH     | Sr SDE   | src/y.rs:12 | ... | Yes |
| 3 | NIT      | QA       | src/z.rs:99 | ... | No (below threshold) |

Fixing N of M findings this iteration.
```

### Step 3: Fix

Implement all fixes at or above the threshold. For each fix:
- State which finding you're addressing
- Make the minimal change needed
- Do not refactor unrelated code

### Step 4: Decide Next Iteration

Count how many findings were fixed this iteration.

- **If 0 fixes were made** → Done. Print summary and stop.
- **If fixes were made AND iteration < 4** → Set `iteration = iteration + 1`, go to Step 1.
- **If iteration = 4** → Done regardless. Print summary and stop.

### Step 5: Summary

When the loop exits, print:

```
## Review Complete

| Iteration | Findings | Fixed | Threshold |
|-----------|----------|-------|-----------|
| 1         | N        | N     | All       |
| 2         | N        | N     | MEDIUM+   |
| ...       | ...      | ...   | ...       |

Total iterations: N
Total findings: N
Total fixed: N
Remaining (below final threshold): N
```

Versions

The database has no other skills named code-review-iterative.

Database snapshot: 13.11.2026 · package version 8a35fc869719 (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

  • 100design-review-iterative
    Run iterative design document review with diminishing fix thresholds. Use when the user says "review my design", "design review", "review this design doc", "review the design iteratively", or after completing a design document.
  • 96iterative-self-review
    Use only when the user explicitly invokes this skill.
  • 100ml-design-review-iterative
    Run iterative ML design review with diminishing fix thresholds. Use when reviewing designs for ML systems, evaluation harnesses, automated search systems, scoring functions, data pipelines, or any design where statistical soundness, data leakage, overfitting risks, and metric…
  • 100code-review
    L2 特性级多维度代码审查。Feature 全部任务完成后执行。 用法:/code-review --base <branch> TRIGGER when: 用户运行 /code-review,或在 merge 前要求特性级代码审查。
  • 100requesting-code-review
    Use when completing tasks, implementing major features, or before merging to verify work meets requirements
  • 100pr-review-team
    Team-based PR review with specialized parallel agents, CI integration, and iterative fix loop. Use when: "review this PR", "PR review", "PRレビュー", "PRをレビューして", "チームでPRレビュー".

Feedback and reports

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