Skills library for AI coding agents — discover, pin, and run reusable capabilities through the Skills CLI and MCP server for Claude Code, Codex CLI, Gemini CLI, OpenCode, and more.

bun install -g @hasna/skills

Requires Bun 1.0+.

# Browse skills interactively
skills

# Hosted setup is the recommended interactive path
skills setup --mode hosted
skills auth login

# Local-only setup stays available and does not require an account
skills setup --mode local

# Optionally pin a skill preference in this project
skills pin image

# Register the Skills MCP server with every supported agent
skills setup agents

# See what a skill needs
skills info image

# Premium skills run through the configured hosted API
skills run image "a cat sitting on a windowsill"

# Free/local skills can still use your own provider keys when documented
skills requires brand-style-guide
OPENAI_API_KEY=... skills run brand-style-guide ./brand-notes.md

Premium skills are hosted SaaS runs. The CLI and MCP server submit them to the configured hosted API, create local run metadata, and then expose status and artifact commands. They do not fall back to bundled local execution when auth is missing or the platform is unavailable.

Use SKILLS_API_KEY or skills auth login for premium hosted execution:

skills setup --mode hosted
skills auth login
skills run image "editorial product photo on a white sweep"
skills runs status <run-id>
skills exports download <run-id>

skills auth login uses browser/device-code auth by default. Email code login is still available with skills auth login --email you@example.com.

SKILLS_API_KEY is the hosted account credential. It is not a provider credential. Provider keys such as OPENAI_API_KEY, ANTHROPIC_API_KEY, or GEMINI_API_KEY remain supported only for free/local OSS skills whose requirements explicitly document local provider use.

• --json — Output as JSON (pipeable)
• --brief — One-line format
• --remote — Read browse/search data from SKILLS_API_URL or config apiUrl
• --dry-run — Preview without applying changes
• --verbose — Debug logging to stderr
• --no-color — Disable ANSI colors
• -o, --overwrite — Refresh existing pin metadata

Commands that support --json write exactly one JSON value to stdout and keep human diagnostics off stdout. Error cases set a non-zero exit code and return an object with an error field where the command shape is not already an array.

Stable command shapes:

• Browse: list, search, categories, tags return arrays.
• Skill details: info, docs, requires, validate, diff, test, doctor, auth, whoami, and outdated return command-specific objects or arrays documented by their field names.
• Project state: pin, unpin, update, init, import, create, and sync return result objects/arrays; --dry-run --json returns { "dryRun": true, "actions": [...] } where applicable.
• Runtime: run --json <skill> ... returns { "skill", "args", "exitCode", "stdout", "stderr", "error", "run" }. Premium remote runs include { "contractVersion": 1, "remote": true, "remoteRun", "pricing", "run", "nextActions" } and return immediately with status commands such as skills runs status <run-id> and skills exports download <run-id>.
• Config and schedules: config * --json and schedule * --json return machine-readable status objects.
• MCP registration: mcp --register <agent> --json returns { "registered": number, "results": [...] }.

Local bundled skills remain the default for discovery. To point browse/search commands at a compatible hosted registry, set an API base URL:

export SKILLS_API_URL=https://skills.md/api/v1
# or persist it:
skills config set apiUrl https://skills.md/api/v1

skills list --remote --json
skills search transcribe --remote --json
skills categories --remote
skills tags --remote --json

If the URL is an origin such as https://skills.md, the CLI requests /api/v1/skills. If it already ends in /api or /api/v1, the CLI appends /skills.

Authenticated remote listing and hosted premium execution use SKILLS_API_KEY or the credential saved by skills auth login.

For the reusable upstream contract, see docs/architecture/reusable-skills-engine.md.

skills mcp    # stdio transport (use with Claude/Codex MCP config)

Long-lived Streamable HTTP transport (default port 8836, bind 127.0.0.1 only):

skills-mcp --http
# or
MCP_HTTP=1 skills-mcp

# override port
skills-mcp --http --port 8836
MCP_HTTP_PORT=8836 skills-mcp --http

Endpoints: GET /health → {"status":"ok","name":"skills"}, MCP at /mcp. Uses stateless StreamableHTTPServerTransport (shared process, many clients). skills-mcp without flags still uses stdio (unchanged).

The MCP server exposes 20+ tools including list_skills, search_skills, pin_skill, unpin_skill, pin_category, list_pinned_skills, get_skill_info, get_skill_docs, get_requirements, run_skill, get_run_status, schedule_skill, detect_project_skills, validate_skill, and more.

skills mcp --register claude    # Auto-register with Claude Code
skills mcp --register all       # Register with all supported agents

skills setup --mode hosted
skills auth login
skills billing status

Hosted account, billing, and credit management use the configured hosted API. The public package only stores local configuration and CLI credentials; Stripe, customer records, and hosted execution remain platform concerns.

src/
├── cli/index.tsx           # Commander.js CLI + Ink TUI
├── mcp/index.ts            # MCP server (stdio) with ~20 tools
├── lib/
│   ├── registry.ts          # 202+ entries, search, categories, tags
│   ├── installer.ts         # Project pins and disabled source-copy paths
│   ├── project-state.ts     # .skills/project.json preferences
│   ├── run-state.ts         # .skills/runs and .skills/exports metadata
│   ├── skillinfo.ts         # Docs, requirements, env/system detection
│   ├── scheduler.ts         # Cron-based skill execution
│   ├── config.ts            # Global + project config loading
│   └── utils.ts             # normalizeSkillName()
├── index.ts                 # Library re-exports (npm package entry)
└── *.test.ts                # Test files

skills/                      # 202+ public skill contracts and local OSS skills
├── _common/                 # Shared utilities
└── */                       # Local skills include src/; hosted skills are metadata-only

Skills are discovered from the remote registry or bundled OSS registry. Project folders and agent-native skill folders are never used as skill libraries.

.skills/ is runtime/output state only:

.skills/
├── project.json              # optional pins/preferences, no secrets
├── runs/YYYY-MM-DD/<run-id>/  # run.json, artifacts.json, events.ndjson, logs/
├── exports/<skill>/<run-id>/  # files produced by a run
└── tmp/

Auth stays global in ~/.hasna/skills/auth.json. Registry and doc caches belong in ~/.cache/skills or the remote API, not inside project .skills.

bun install
bun run build              # Build CLI, MCP, library, and types
bun run dev                # Run CLI in dev mode (no build needed)
bun test                   # Run all tests
bun run typecheck          # TypeScript type checking

1. Create skills/{name}/ with src/index.ts, package.json, tsconfig.json, SKILL.md
2. Add an entry to the SKILLS array in src/lib/registry.ts
3. Run skills validate <name> --json to check package metadata, bin entries, docs, and SKILL.md frontmatter
4. Run bun test to verify registry-wide validation passes

Premium hosted skills should add public contracts, pricing, docs, and tests without adding private provider routing, hosted worker code, or secrets to the OSS package.

Custom skill directories are auto-discovered from ~/.hasna/skills/custom/. Project .skills/ is reserved for runtime state and outputs.

Global configuration is stored in ~/.hasna/skills/. Auth is stored in ~/.hasna/skills/auth.json. Project runtime data is stored in .skills/runs, .skills/exports, .skills/tmp, and optional .skills/project.json.

Apache-2.0 — see LICENSE