ancienttwo/repo-harness/repo-harness-setup

repo-harness-setup

Worker entrypoint. Use the Bot-assigned mode and target repo. If either is missing, return the decision to the Bot.

Load only the reference for the assigned mode.

Shared Preflight

1. Confirm the target repo path (pwd, or an explicit --repo argument).
2. Inspect the proposed changes with repo-harness init --repo <repo> --dry-run (not ci-scope).

Mode Selection

• No harness yet, or refreshing an existing install -> references/init.md.
• Inspector reports legacy docs or stale harness artifacts -> references/migrate.md.
• Harness present, needs latest contract/helpers/templates -> references/upgrade.md.
• A specific workflow surface is broken -> references/repair.md.
• New project/app/module skeleton, no existing repo workflow -> references/scaffold.md.
• Add or sync one capability boundary only -> references/capability.md.
• Impact-scoped PR CI with a release-only full matrix -> references/ci-scope.md.

Boundaries

• Never write user-level (HOME) state from a repo-scoped mode; user-level setup is the separate repo-harness update command.
• Preserve user-authored repo files unless the workflow contract owns the generated surface; remove only ownership=known_generated files.
• Does not create an application stack from any mode except scaffold.
• Does not expose internal helper scripts (create-project-dirs, direct shell init adapters, hooks-init, docs-init) as public commands.
• Does not infer capability prefixes from broad directory globs; always use explicit prefixes.
• Return a request spanning multiple modes to the Bot for scope and order.