jewzaam-reviews

jewzaam-reviews

A Claude Code plugin bundling a connected pipeline of review skills. Producer skills emit Findings-<skill-name>[-<scope>].json (validated against schemas/findings.schema.json) plus a script-rendered .md companion; apply-review consumes the JSON, applies each finding as an isolated commit, and emits Report-apply-review.json summarizing what happened.

Skills

Skill	Output files	Description
jewzaam-reviews:review	Findings-review[-<scope>].{json,md,-supplementary.md}	Multi-agent codebase review across parallel dimensions
jewzaam-reviews:standards	Findings-standards.{json,md,-supplementary.md}	Audit repos against ~/source/standards/ personal standards library
jewzaam-reviews:update-pr	Findings-update-pr-<number>.{json,md}	Fetch GitHub PR review comments and supplementary feedback
jewzaam-reviews:c4-reverse-engineer	Findings-c4-reverse-engineer.{json,md}	Reverse-engineer C4 architecture diagrams and behavioral spec from a codebase
jewzaam-reviews:apply-review	Report-apply-review.json	Consume any Findings-*.json, apply as one-commit-per-finding, emit an action report

Installation

/plugin marketplace add jewzaam/jewzaam-reviews
/plugin install jewzaam-reviews@jewzaam-reviews-marketplace

Permissions

Skills invoke Python and Bash scripts from the plugin cache. To avoid repeated permission prompts, add these to your global (~/.claude/settings.json) or project (.claude/settings.json) allowlist:

{
  "permissions": {
    "allow": [
      "Bash(bash ~/.claude/plugins/cache/jewzaam-reviews-marketplace/**)",
      "Bash(python ~/.claude/plugins/cache/jewzaam-reviews-marketplace/**)",
      "Bash(python3 ~/.claude/plugins/cache/jewzaam-reviews-marketplace/**)",
      "Bash(~/.claude/plugins/cache/jewzaam-reviews-marketplace/**)",
      "Read(~/.claude/plugins/cache/jewzaam-reviews-marketplace/**)",
      "Bash(bash ~/.claude/plugins/marketplaces/jewzaam-reviews-marketplace/**)",
      "Bash(python ~/.claude/plugins/marketplaces/jewzaam-reviews-marketplace/**)",
      "Bash(python3 ~/.claude/plugins/marketplaces/jewzaam-reviews-marketplace/**)",
      "Bash(~/.claude/plugins/marketplaces/jewzaam-reviews-marketplace/**)",
      "Read(~/.claude/plugins/marketplaces/jewzaam-reviews-marketplace/**)"
    ]
  }
}

Pipeline

graph LR
    review[review] -->|Findings-review.json| apply[apply-review]
    standards[standards] -->|Findings-standards.json| apply
    updatepr[update-pr] -->|Findings-update-pr-N.json| apply
    c4[c4-reverse-engineer] -->|Findings-c4-reverse-engineer.json| apply
    apply -->|Report-apply-review.json| done((done))

Usage

From any project repo:

/jewzaam-reviews:review                 # Multi-dimensional codebase review
/jewzaam-reviews:standards              # Audit against ~/source/standards/
/jewzaam-reviews:update-pr              # Pull PR review comments
/jewzaam-reviews:c4-reverse-engineer    # Generate C4 diagrams + spec
/jewzaam-reviews:apply-review           # Apply any Findings-*.json

Filename convention

Two document types, two prefixes:

• Findings-<skill-name>[-<scope>].{json,md[,-supplementary.md]} — produced by the four producer skills. "Findings" because these are things the reviewer found in the user's code.
• Report-apply-review.json — produced by apply-review. "Report" because it summarizes actions taken, not findings. No markdown (no user-facing review document).

Filenames carry the skill name, not the project name. Project identity lives inside the JSON envelope's project.name field (the working directory is the project, so repeating it in every filename was redundant). Scope suffixes are used when a skill supports multiple scoped runs (e.g., PR numbers for update-pr, scope slugs for review).

Development

make help               # Show all targets
make check              # Run all checks (test + version-check)
make test               # Run pytest across plugin + skills
make version-check      # Validate semver consistency
make version-bump-patch # Bump patch (e.g. 0.2.8 → 0.2.9)
make version-bump-minor # Bump minor (e.g. 0.2.8 → 0.3.0)
make version-bump-major # Bump major (e.g. 0.2.8 → 1.0.0)

Shared handoff schema

All producer and consumer skills validate their JSON against schemas/findings.schema.json. The schema discriminates on a top-level source field (review / standards / c4-reverse-engineer / apply-review) and carries a uniform issues[] array for meta-issues from the run. update-pr is absent from the enum by design — it emits review-shaped findings with optional pr_comment fields, under source: "review". See CLAUDE.md for the invariants and resources/handoff-contract.md for the full contract.

License

Apache-2.0