Agent configs scatter across repos and silently diverge. Skilldeck keeps one local skill library and deploys to each tool in its native format so you don't rebuild behavior for every repo. Repo: https://github.com/ali-erfan-dev/skilldeck
It maps formats: .claude/skills, .cursor/rules/*.mdc, AGENTS.md, .windsurfrules, etc. Ten built‑in targets (Claude Code, Cursor, Copilot, Windsurf, Codex). Local‑only (Win/mac/Linux). Drift detection flags deployed skills that fall out of sync.
The app was built by Claude Code with a harness: ground‑truth JSON, Playwright E2E checks, a regression gate using a surfaces map, and a feature intake protocol. 31 features shipped across autonomous sessions. Agent coding needs engineering rigs, not improvisation.
Takeaway for agencies, advisors, lawyers: treat agent behavior files like code—version, test, audit. Skilldeck gives repeatability, pull‑back edits, and compliance signals. Have you used a local skill library to standardize agent behavior across client repos?
Top comments (0)