# Code Review > Run the quality-triage code-review agent (ML smells, Python smells, technical-debt classification, AST code intelligence) on a path or GitHub URL. Use when the user asks for a code review, quality audit, smell detection, complexity hotspots, or technical-debt analysis of Python code. Use when this capability is needed. - Skill: `tomevault-io/code-review-294` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/code-review-294` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/code-review-294/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/code-review-294 --- # code-review skill This skill drives the `code-review` CLI / `code-review-mcp` MCP server from the `quality-triage` project. The canonical contract is `skills.md` at the repo root. ## When to use - "Review this project / file", "find code smells", "find ML anti-patterns", "what's the technical debt here", "show complexity hotspots". ## Setup (once) ```bash uv sync # core uv sync --extra mcp # if using the MCP server ``` ## Preferred: MCP tools If the `code-review` MCP server is connected, call its tools directly: `detect_python_smells`, `detect_ml_smells`, `analyze_code_intelligence`, `classify_technical_debt`, `list_python_files`, `read_file`. ## CLI fallback ```bash code-review review --output reports/review.md code-review run-tool python-smells --type all --format json code-review run-tool ml-smells --format json code-review run-tool code-intel --format json code-review run-tool classify-td --text "TODO: ..." --category security code-review doctor # verify environment / backends code-review providers # show configured LLM providers ``` ## Output contract 1. Executive summary 2. Critical issues first, with `file:line:col` 3. ML-specific issues 4. Code quality & architecture issues 5. Technical debt categories 6. now/next/later remediation plan Never invent commands, files, or tool names. Prefer the smallest command that answers the question; escalate to a full `review` only when needed. --- > Source: [KarthikShivasankar/quality-triage](https://github.com/KarthikShivasankar/quality-triage) — distributed by [TomeVault](https://tomevault.io).