analyze

Analyze the current project to prepare for agent team creation. Read everything relevant and produce a structured analysis.

Step 1: Project Foundation

Read these files if they exist:

• README.md, CONTRIBUTING.md, CHANGELOG.md
• .github/pull_request_template.md
• .github/ISSUE_TEMPLATE/

Step 2: Tech Stack

Identify:

• Language and version (go.mod, package.json, Cargo.toml, pyproject.toml, etc.)
• Build system and commands
• Linter config with exact rules
• Test framework, assertion library, coverage requirements
• CI/CD workflows (.github/workflows/)

Step 3: Code Patterns

• Map directory structure with Glob
• Read 5-10 key source files for patterns
• Read 5-10 test files for testing conventions
• Identify the extension/plugin/registration pattern (how new features are added)

Step 4: PR Review Analysis

Find the lead reviewer:

gh pr list --state all --limit 100 --json number,reviews --jq '.[].reviews[].author.login' 2>/dev/null | sort | uniq -c | sort -rn | head -5

Find external contributor PRs:

gh pr list --state all --limit 200 --json number,title,state,author --jq '.[] | "\(.number) [\(.state)] \(.author.login): \(.title)"' 2>/dev/null

For each external PR with lead reviewer feedback, extract BOTH:

PR-level reviews:

gh api repos/OWNER/REPO/pulls/NUMBER/reviews --jq '.[] | select(.user.login == "REVIEWER") | .body' 2>/dev/null

Inline code comments:

gh api repos/OWNER/REPO/pulls/NUMBER/comments --jq '.[] | select(.user.login == "REVIEWER") | "FILE: \(.path):\(.line // .original_line)\nCOMMENT: \(.body[:400])"' 2>/dev/null

Analyze at least 15-20 PRs. Rank reviewer patterns by frequency into tiers.

Step 5: Context Folder

Read everything in .context/ if it exists.

Output

Save the analysis to .context/forge-analysis.md with these sections:

# Forge Analysis: [Project Name]

## Tech Stack
[language, build, linter, test framework, CI]

## Code Patterns
[file structure, naming conventions, registration pattern]

## Test Conventions
[package naming, assertion library, fixture patterns, coverage requirement]

## PR Review Patterns
### Lead Reviewer: [name]
### Tier 1 (50%+ of PRs)
[patterns with exact quotes]
### Tier 2 (25-50%)
[patterns with exact quotes]
### Tier 3 (10-25%)
[patterns with exact quotes]
### Things NOT to Flag
[verified false positives]

## PR Format
[title format, body structure, checklist]

## Contribution Type
[what external contributors typically add]

## Build/Generate Commands
[commands that must run before submitting]

Tell the user the analysis is complete and they can run /claude-army:create-team next.