Skill files (.claude/skills/, .cursor/rules/*.mdc,
AGENTS.md, .windsurfrules) are becoming a core part of
AI-assisted development workflows. The problem: they
scatter across projects, diverge silently, and every new
repo means rebuilding behavioral config from scratch. Each
tool uses a different format and location.
Skilldeck keeps one local library and deploys to any tool
in the correct format automatically. Ten built-in target
profiles cover Claude Code, Cursor, Copilot, Windsurf,
Codex, and more. Drift detection shows when a deployed
skill has fallen out of sync with the library version.
Bidirectional sync lets you pull improvements back from a
project into the library.
The interesting part technically: the entire app was built
by Claude Code using a harness engineering methodology —
a ground truth JSON file, Playwright E2E verification
tests, a regression gate based on a surfaces map, and a
feature intake protocol. 31 features across multiple
autonomous sessions with no manual application code. I
wrote two articles about the harness approach if that side
is interesting.
No cloud, no backend, local filesystem only.
Windows/macOS/Linux. Open source.
github.com/ali-erfan-dev/skilldeck