Skip to content

v0.1.0 — SkillFoundry launch

Latest
Latest

Choose a tag to compare

View all tags
@NSBen NSBen released this 23 Jul 09:10
· 3 commits to main since this release
v0.1.0
8ab0cfb

SkillFoundry v0.1.0 — 🔥 Launch

The open, vendor-neutral toolchain for authoring, validating, testing & publishing AI agent skills — one SKILL.md spec, one CLI, every major coding agent.

Why this exists

Claude Code, Cursor, Codex, OpenCode and Cline all now support "skills" (reusable capability packs). But the ecosystem is fragmented:

  • most offerings are static collections with no quality gate,
  • many are vendor-locked to one client,
  • there is no shared spec and no tooling to keep skills correct.

SkillFoundry fills that gap: an open Markdown-based spec + a single CLI that works across all five clients.

What's in v0.1.0

  • forge CLI with 9 commands: init · validate · test · doctor · search · add · list · publish · schema
  • Open spec — SKILL.md + exported skill.schema.json (JSON Schema), vendor-neutral
  • 15 ready-to-run starter skills (security-audit, write-tests, pr-review, sql-optimize, i18n, explain-codebase, doc-gen, commit-msg, dependency-upgrade, api-client-gen, refactor-smell, changelog, test-data-gen, explain-error, release-notes) — all pass validate + test
  • Cross-client install — detect & install to Claude Code / Cursor / Codex / OpenCode / Cline from one command
  • Offline test runner — structural tests with no API keys
  • Full docs, 27 passing tests, ruff clean, CI + registry-test workflows

Install

pip install skillfoundry

Quickstart

forge init my-skill && cd my-skill
forge validate .        # rule-based checks (--strict promotes warnings to errors)
forge test .            # offline structural tests
forge doctor            # environment health check
forge add security-audit   # pull a starter skill into your project

Highlights

  • Vendor-neutral: the same skill file installs into any supported client.
  • Standard-first: generate a JSON Schema from any spec and validate against it.
  • Safe by default: validate + test catch missing frontmatter, broken names, empty bodies, bad tags before you ship.
  • MIT licensed, contributions welcome (see CONTRIBUTING.md).

Links

  • Spec → docs/spec.md
  • Authoring guide → docs/authoring.md
  • Registry model → docs/registry.md

Give it a ⭐ if a neutral skills toolchain is something you've wanted. Feedback and PRs are very welcome!

Assets 2
Loading