Code Review
Reads the project type from CLAUDE.md, then loads the appropriate language-specific review checklist. Universal review principles apply to all types.
Step 0 — Load universal principles
Load ~/.hortora/garden/approaches/code-review.md before proceeding.
All review severity models, reporting formats, and workflow steps are defined there.
Step 1 — Detect project type
grep -A 2 "## Project Type" CLAUDE.md 2>/dev/null | grep "^type:"
Extract: java | ts | python | generic
If type is missing or generic, inspect staged files:
.javafiles orpom.xmlchanged → treat asjava.ts/.tsxfiles changed → treat asts.pyfiles changed → treat aspython- Mixed → flag and ask user which review to run
Step 2 — Load language-specific checklist
| Project type | File to read |
|---|---|
java |
~/.claude/skills/code-review/java.md |
ts |
~/.claude/skills/code-review/typescript.md |
python |
~/.claude/skills/code-review/python.md |
Read the file, then execute the review workflow it describes.
Skill Chaining
Invoked by: [java-dev], [ts-dev], [python-dev] before committing; [java-git-commit] when no review has been run this session
Invokes: [security-audit] for auth/payment/PII code (offered, not automatic); [git-commit] after approval if user wants to commit
Source: mdproctor/cc-praxis — distributed by TomeVault.