Normalized repo-owned guidance for agents. Keep it aligned with docs/references/agent-dev-workflow.md and .rabbit/context.yaml.
- context: ./.rabbit/context.yaml
- workflow_ref: ./docs/references/agent-dev-workflow.md
- Make sure
dev.kititself is current, then rundev.kit. - Read
.rabbit/context.yamlfirst when it exists. - Read the highest-priority refs and manifests that
context.yamlpoints to. - Run
dev.kit repoafter repo changes or when context is missing or stale.
- treat
.rabbit/context.yamlas generated repo evidence, not hand-authored guidance - prefer repo-declared commands and manifests over guessed behavior
- keep durable workflow guidance in repo-owned docs, especially
docs/references/agent-dev-workflow.md - when a gap appears, repair the owning repo asset and rerun
dev.kit repo - use live GitHub data only when the task needs issues, PRs, reviews, workflow runs, or alerts
- read:
README.md,changes.md,.github/workflows/,docs/ - verify:
make test
AGENTS.mdis optional and repo-owned in this repo- the reference doc is the fuller source of examples, Q&A, and best practices