/agentic-skills-init

Marry a target repo to agentic-workflow-skills. Every other agentic-workflow-skills skill reads this repo's config — its stack, its tracker, its UI gates, its data defaults, its documents config — and none of them may guess. This skill is where those facts get pinned, once, from evidence.

You are writing config for a repo you may never have seen. Read it first. Nothing in the output may be assumed: every stack fact is either backed by a file in the repo or recorded as "none detected".

Output

Five markdown files under docs/agentic-skills/ in the target repo's root:

• stack.md — language, framework, ORM, database, test runner, each with its evidence
• tracker.md — where issues live, how tracker operations map, the triage label strings
• ui-gates.md — the six hard gates, the repo's paths floor, and its gate exceptions
• data.md — fabrication defaults: profiles, volumes, locale blend, seed, manifest and row-marker conventions
• documents.md — the document set's prose language and doc root, with the rule that a doc root without an activity list is no document set at all

Re-runs update these files in place. Sections a human has curated — gate exceptions, the paths floor, label overrides, the documents config's two pinned lines — are preserved, not rewritten from templates.

Process