What problem does it solve? Repositories using the sassy-dog workflow skills need per-repo configuration (.claude/sassy-dog/*.md plus .claude/settings.json plugin declarations), and hand-writing or migrating that config from older generated skills is error-prone: facts drift, merge policies go stale, and legacy skill directories must be removed safely without losing project-specific prose. ## Core Features & Use Cases - Four operating modes: migrate (extract facts and prose from legacy generated skills, re-verify against live state, then delete old directories), update (re-verify every configured fact against live GitHub state while carrying user prose verbatim), adopt (fold hand-written skills into config), and create (render fresh config from templates). - Capability detection and interview: runs scripts/detect-capabilities.sh to probe merge queues, boards, CI workflows, migrations, codegen, Sentry, and secret managers, then confirms policy decisions through a structured interview. - Safe writes: prints every file in full and writes only after user approval, never deletes directories lacking a generated-by: marker, and verifies settings.json declares both the marketplace and the plugin. - Use Case: A repo still carrying old .claude/skills/plate-it/ generated skills is migrated to .claude/sassy-dog/survey-work.md config, with its merge-queue setting re-verified via GraphQL before the legacy directory is deleted. ## Quick Start Ask the agent to set up this repo's sassy-dog workflow config by running the setup-config skill from inside the target repository.