Quickstart (River Review)

This guide covers the minimal configuration to get River Review running. River Review turns tacit knowledge into versioned, repo-owned Skills (a Skill Registry) shared as a team asset, and runs on three axes: a capability pack (usually no LLM key needed), review skills, and a review team of perspective-based reviewers. The headless GitHub Action this guide sets up is one path for calling those three axes from CI.

1. Prepare skills/{upstream,midstream,downstream} in your repository and create skills using Markdown + YAML frontmatter (Refer to schemas/skill.schema.json). See Upstream / Midstream / Downstream phases for an explanation of each phase.
2. Add a GitHub Actions workflow to .github/workflows/river-review.yml or similar.
3. Configure an LLM API key (OPENAI_API_KEY / ANTHROPIC_API_KEY / GOOGLE_API_KEY) in your repository Secrets for the GitHub Action. Additional tokens are only needed for custom integrations or extensions. Note that an LLM key is required only for this headless GitHub Action (or the standalone CLI) — normal use, where an AI agent (Claude Code / Cursor / …) applies the skills, needs no key (see What is River Review § Execution model).
4. Create a PR and verify that skills for the specified phase are executed.

Refer to the root README.md for a minimal Actions definition sample.