Agent Skills: Best Practices Extractor

Extract and generate coding best practices from PR review comments. Use when the user asks to "extract best practices", "analyze PR comments", "generate coding standards", "create best practices from PRs", or "update coding guidelines from reviews".

UncategorizedID: ItamarZand88/claude-code-agentic-engineering/best-practices-extractor

Install this agent skill to your local

pnpm dlx add-skill https://github.com/ItamarZand88/claude-code-agentic-engineering/tree/HEAD/plugins/workflow/skills/best-practices-extractor

Skill Files

Browse the full folder contents for best-practices-extractor.

Download Skill

Loading file tree…

plugins/workflow/skills/best-practices-extractor/SKILL.md

Skill Metadata

Name
best-practices-extractor
Description
Extract and generate coding best practices from PR review comments. Use when the user asks to "extract best practices", "analyze PR comments", "generate coding standards", "create best practices from PRs", or "update coding guidelines from reviews".

Best Practices Extractor

Extract coding best practices from PR review comments and codebase analysis to generate .claude/best-practices/ documentation.

When to Use

  • Initial setup of coding standards for a project
  • Regular updates to capture new team feedback
  • Analyzing PR comments to document patterns
  • Generating best practices from code review history

Extraction Methods

All outputs are saved to .claude/pr-review-comments/.

Method 1: Incremental Update (Recommended)

For regular updates - fast and tracks state:

cd plugin/skills/best-practices-extractor
bash scripts/incremental_update.sh OWNER REPO_NAME

First run does full extraction; subsequent runs fetch only new PRs.

Output: .claude/pr-review-comments/{repo}_inline_comments.ndjson

Method 2: Full Extraction

For first-time setup or comprehensive analysis:

bash scripts/extract_pr_comments.sh REPO_NAME

Output: .claude/pr-review-comments/{repo}_inline_comments.ndjson

Method 3: Sort by File Tree

Organize extracted comments by directory structure:

bash scripts/sort_comments_by_filetree.sh .claude/pr-review-comments/{repo}_inline_comments.ndjson

Output: .claude/pr-review-comments/sorted/

Generating Best Practices

After extracting comments:

  1. Identify patterns - Look for recurring feedback themes
  2. Categorize - Group by type (naming, error-handling, type-safety, etc.)
  3. Create documentation - Generate .claude/best-practices/ files

Output Structure

.claude/best-practices/
├── README.md              # Index
├── naming-conventions.md
├── error-handling.md
├── type-safety.md
├── testing.md
└── ...

Guideline Format

Each guideline should include:

## [Number]. [Title]

**Guideline:** [Clear statement]

**Why:** [Impact explanation]

**Example:**
```typescript
// Good
{example}

// Bad
{counter_example}

Source: PR #{number}


## Prerequisites

- **GitHub CLI (`gh`)**: Authenticated - check with `gh auth status`
- **jq**: JSON processor - check with `jq --version`

## Integration

This skill **generates** best practices. The `code-reviewer` agent and `/4_review` command **validate** against them.

Workflow:
1. Extract PR comments (this skill)
2. Analyze and generate `.claude/best-practices/`
3. Reviews automatically validate against best practices

## Additional Resources

- **`references/default-categories.md`** - Example category definitions
- **`references/scripts-guide.md`** - Detailed script documentation