AgentFlow
Contributing

Contributing

How to contribute to AgentFlow and keep docs in sync with the CLI.

Contributing

The living checklist, review expectations, and repository policies live in CONTRIBUTING.md at the repo root — start there for anything beyond this short map.

Quick checklist

Before you open a PR, load project context, align scope with active handoff and product specs, prove the Go tree still builds and tests clean, regenerate CLI reference when commands move, run the docs site build for MDX edits, and keep each PR focused without leaking secrets.

  1. Read AGENTS.md and docs/ai/context-map.md
  2. Align features with docs/ai/active/handoff.md and product specs (spec-postv123.md, etc.)
  3. make build && go test -race ./...
  4. Update CLI docs when commands change:
go run ./application/cmd/agentflow docs generate-cli \
  --output docs-site/content/docs/en/cli/generated
  1. cd docs-site && npm run build for MDX changes
  2. One PR per coherent topic; no secrets in commits

Documentation

Public prose targets docs-site/content/docs/ (English site uses Fumadocs). Internal canon and steering notes stay under docs/ai/. When code and docs disagree, problems.md at the root tracks known drift.

Architecture decisions

When you change a durable tradeoff, record it where future contributors look: docs/ai/05-decisions.md plus structured entries in docs/decisions/ if your team uses that folder layout.