Process Code Review

Overview

Fix or explicitly defer findings from a Monty review document, interactively or with --auto.

This skill ships inside the Process Code Review plugin and can be installed through the Claude Code marketplace or directly in Codex from its skill path.

Parent Surface

When to Use This Skill

This skill is the second step in the code review workflow:

Before applying fixes, read local typing policy docs when present (for example docs/python-typing-3.14-best-practices.md, TY_MIGRATION_GUIDE.md, AGENTS.md) and align type-check behavior with those rules.

• After running /monty-code-review:code-review to generate a review document.
• When you have a *_review.md file with severity-tagged issues to process.
• When you want systematic, tracked resolution of code review findings.
• When reviewing a colleague's code review document and fixing issues one-by-one.

Step 1: /monty-code-review:code-review → Creates *_review.md
Step 2: /process-code-review:process-review *_review.md → Fix/skip issues
Step 3: /backend-atomic-commit:pre-commit → Run pre-commit checks and fix issues

Example Prompts

• "Process the wellbeing_service_review.md file and help me fix the issues."
• "Use your process-code-review skill on optimo_core/docs/code_reviews/api_review.md"
• "Go through the code review findings in this file and fix or skip each one."
• /process-code-review:process-review myfile_review.md - Interactive mode (default)
• /process-code-review:process-review --dry-run myfile_review.md - Preview only, no changes
• /process-code-review:process-review --auto myfile_review.md - Auto-fix all issues
• /process-code-review:process-review --auto --severity=NIT myfile_review.md - Auto-fix only NITs
• /process-code-review:process-review --severity=BLOCKING api_review.md - Process only BLOCKING issues

Processing Workflow

When this skill is active and you are asked to process a review document:

Read the review file and extract all issues. Look for:

Common patterns to look for:

Use TodoWrite to track all issues:

Example todo list structure:

For each issue, starting with highest severity, display:

In default interactive mode, ask Fix this issue or skip it? and wait. With --auto, continue through the selected findings without per-issue approval; stop for consequential ambiguity or actions outside the authorized scope. --dry-run only reports proposals and never changes source or review status.

formatter/linter pass solely because the next section names it again.

After each fix, run quality checks:

For touched files, "baseline acceptable" is not allowed. Resolve all active type-gate errors before moving on. If repo policy requires wider type checks before merge, run them before final completion.

Fix check failures caused by the change and required touched-file gate errors. Report unrelated failures without expanding the task. Reuse passing checks for unchanged inputs; rerun affected checks after edits and any required wider gates before completion.

When all issues are processed, provide a summary:

• Severity tags: [BLOCKING], [SHOULD_FIX], [NIT]
• Issue titles: Following the severity tag
• Location: File path and line numbers
• Code snippets: Current problematic code
• Problem description: Explanation of the issue
• Proposed fix: Suggested code or approach
• Group by severity: BLOCKING first, then SHOULD_FIX, then NIT
• Mark each as pending initially
• Update status as you process each issue
• ruff check --fix - Auto-fix linting issues
• ruff format - Format code
• Active type gate (ty if configured; else pyright; else mypy) - Type check
• python manage.py check - Django system checks (if applicable)

### [BLOCKING] Issue Title

**Location:** path/to/file.py:L120-L145

**Current code:**
...

**Problem:**
...

**Proposed fix:**
...

[BLOCKING] C1. N+1 Query Pattern - pending
[BLOCKING] C2. Missing Org Scoping - pending
[SHOULD_FIX] S1. Performance Issue - pending
[NIT] N1. Docstring Missing - pending

Status Markers

Use these markers when updating review documents:

Always include the date: Status: ✅ FIXED (2025-01-15)

Resources

Declared allowed tools:

Installation

Switch between Claude Code and Codex, then copy the install command for the runtime you use.

claude plugin marketplace add DiversioTeam/agent-skills-marketplace
claude plugin install process-code-review@diversiotech

CODEX_HOME="${CODEX_HOME:-$HOME/.codex}"
python3 "$CODEX_HOME/skills/.system/skill-installer/scripts/install-skill-from-github.py" \
  --repo DiversioTeam/agent-skills-marketplace \
  --path plugins/process-code-review/skills/process-code-review

Invocation:

/process-code-review:process-review

name: process-code-review