skillslap-mcp

MCP server for SkillSlap — the Slap Stack for AI agent skills. Search, install, compose, and publish skills from your AI coding agent.

Setup

Claude Code

claude mcp add skillslap -- npx skillslap-mcp

Or with an API token (required for publishing, slapping, and agent tracking):

claude mcp add skillslap -e SKILLSLAP_TOKEN=sk_live_... -- npx skillslap-mcp

Cursor / Cline

Add to .cursor/mcp.json (or your MCP settings):

{
  "mcpServers": {
    "skillslap": {
      "command": "npx",
      "args": ["skillslap-mcp"],
      "env": {
        "SKILLSLAP_TOKEN": "sk_live_..."
      }
    }
  }
}

CLI Login (no copy-paste)

Instead of manually copying a token, run:

npx skillslap-mcp login

This opens your browser, you click Authorize, and the token is saved automatically to ~/.config/skillslap/credentials. The MCP server picks it up on next start — no SKILLSLAP_TOKEN env var needed.

Environment Variables

| Variable | Required | Default | Description | |----------|----------|---------|-------------| | SKILLSLAP_TOKEN | For auth'd operations | — | API token from your SkillSlap dashboard (Settings → API Tokens) | | SKILLSLAP_AGENT_NAME | No | hostname | Display name for this agent in the dashboard | | SKILLSLAP_MODEL | No | — | Model hint shown in the agent dashboard (e.g. claude-sonnet-4-5) | | SKILLSLAP_ENV | No | prod | Target environment: local | dev | prod | | SKILLSLAP_API_URL | No | (resolved from env) | Explicit API URL override (highest priority) | | SKILLSLAP_DEV_URL | When SKILLSLAP_ENV=dev | — | Staging / preview deployment URL | | SKILLSLAP_SERVICE_ROLE_KEY | Local/dev admin tools | — | Supabase service role key (for verify_skill, verify_all_unverified) |

Environment resolution order

1. SKILLSLAP_API_URL — explicit override (takes precedence over everything)
2. SKILLSLAP_ENV=local → http://127.0.0.1:3000 (Next.js dev server)
3. SKILLSLAP_ENV=dev → SKILLSLAP_DEV_URL
4. SKILLSLAP_ENV=prod (default) → https://skillslap.com

Agent Tracking

Every agent installation gets a persistent fingerprint stored at ~/.config/skillslap/device_id. This lets the SkillSlap dashboard show which agents are connected to your account, what they've been doing, and their personal skill libraries — without tying identity to a specific token.

You can view and manage your agents at skillslap.com/dashboard/agents.

Tools

search_skills

Search the Slap Stack for agent skills. Returns compact results with install URLs.

• query (string) — Search query (omit to browse all)
• tags (string) — Comma-separated tag filter
• invocation_type (string) — agent | user | tool | context
• sort (string) — relevance | trending | recent | slaps | focus_pool
• limit (number) — Max results (default: 10, max: 50)
• verified (boolean) — Only return verified skills
• username (string) — Filter to skills by a specific author

get_skill

Get details for a specific skill by ID with tiered data loading.

• id (string, required) — Skill UUID
• tier (number) — 1 = metadata, 2 = full (default), 3 = with files

install_skill

Download a skill formatted for your agent, ready to save.

• id (string, required) — Skill UUID
• agent (string) — Agent name: claude-code, cursor, cline, codex, windsurf, aider, copilot, gemini-cli, vscode

list_versions

Get version history for a skill with change summaries.

• id (string, required) — Skill UUID

publish_skill

Publish a new skill to SkillSlap. Requires SKILLSLAP_TOKEN.

• title (string, required) — Skill title (1-100 chars)
• content (string, required) — Skill content in Markdown
• description (string) — Short description (max 500 chars)
• tags (string[]) — Tags (max 10)
• invocation_type (string) — agent | user | tool | context
• status (string) — draft | active

update_skill

Update a skill you own. Requires SKILLSLAP_TOKEN.

• id (string, required) — Skill UUID
• title (string) — New title
• content (string) — New content in Markdown
• description (string) — New description
• tags (string[]) — New tag list (replaces existing)
• status (string) — draft | active | archived
• version (string) — New version string (e.g. "1.1.0")
• change_summary (string) — What changed (stored in version history)

slap_skill

Add a skill to this agent's personal library. Separate from your human slap count — this is a per-agent bookmark for quick access. Requires SKILLSLAP_TOKEN.

• skill_id (string, required) — Skill UUID to add

unslap_skill

Remove a skill from this agent's personal library.

• skill_id (string, required) — Skill UUID to remove

list_agent_slaps

List all skills in this agent's personal library. Each agent has its own independent library — useful for giving different agents access to different skill subsets. Requires SKILLSLAP_TOKEN.

list_my_slaps

List all skills the authenticated user has slapped (human endorsements). Requires SKILLSLAP_TOKEN.

compose_skills

Compose multiple skills into a CLAUDE.md + agent_docs structure. No auth required for public skills.

• skill_ids (string[], required) — Array of skill UUIDs to compose
• project_name (string) — Optional project name for the output

generate_skill

Generate a skill from a natural language description using AI. Requires SKILLSLAP_TOKEN and a BYOK Anthropic API key configured on your account.

• prompt (string, required) — Description of the skill to generate (10-2000 chars)
• auto_publish (boolean) — Publish immediately (default: false, returns draft for review)

verify_skill

Trigger server-side verification for a skill. Runs classify → malware scan → AI analysis. Requires SKILLSLAP_TOKEN or SKILLSLAP_SERVICE_ROLE_KEY (local/dev).

• id (string, required) — Skill UUID to verify

verify_all_unverified

Find all active skills missing a system-tier verification and verify them. Ideal for seeding a fresh local or dev environment. Requires SKILLSLAP_TOKEN or SKILLSLAP_SERVICE_ROLE_KEY.

• limit (number) — Max skills to process (default: 50)
• dry_run (boolean) — List without verifying (default: false)

list_collections

List your skill collections. Returns titles, descriptions, visibility, and the skill IDs in each. Requires SKILLSLAP_TOKEN.

create_collection

Create a new skill collection. Requires SKILLSLAP_TOKEN.

• title (string, required) — Collection title (1-100 chars)
• description (string) — Optional description
• is_public (boolean) — Whether the collection is publicly visible (default: false)

add_to_collection

Add a skill to one of your collections. Requires SKILLSLAP_TOKEN.

• collection_id (string, required) — Collection UUID
• skill_id (string, required) — Skill UUID to add

list_skill_queue

List pending skill generation requests queued by the authenticated user. Use this to check what requests are waiting to be processed. Requires SKILLSLAP_TOKEN.

claim_queue_item

Atomically claim a skill queue item so no other agent processes it. Call this before starting research or skill creation. Requires SKILLSLAP_TOKEN.

• id (string, required) — Queue item UUID (from list_skill_queue)

complete_queue_item

Mark a skill queue item as completed (with the resulting skill ID) or failed (with an error message). Triggers a notification to the user. Requires SKILLSLAP_TOKEN.

• id (string, required) — Queue item UUID
• skill_id (string) — UUID of the published skill (provide on success)
• error (string) — Error message (provide instead of skill_id on failure)

list_webhooks

List all webhooks registered on your account. Returns URL, events, active status, and creation date. Requires SKILLSLAP_TOKEN.

create_webhook

Register a new webhook endpoint. SkillSlap will POST a signed payload to your URL when the specified events occur. Requires SKILLSLAP_TOKEN.

• url (string, required) — Endpoint URL (must be https://)
• events (string[], required) — Event types: skill_published, skill_verified, skill_slapped, skill_installed, queue_completed, queue_failed, agent_session_started

CLI

The package also installs a skillslap CLI for use in terminals and scripts.

# Authentication
skillslap login                                   # browser-based login, saves token automatically

# Discovery
skillslap search "code review"
skillslap search testing --verified --limit=5
skillslap install <skill-id>
skillslap install <username>/<slug>               # install by slug
skillslap install <skill-id> --agent=claude-code

# Publishing
skillslap list
skillslap publish ./SKILL.md --tags="python,testing" --status=active

# Target a specific environment per-command:
skillslap --env=local search "security"
skillslap --env=dev publish ./SKILL.md --status=draft

Local Development Mode

Point the MCP server at your local Supabase instance instead of production. The local service role key is printed by npx supabase status.

{
  "mcpServers": {
    "skillslap-local": {
      "command": "node",
      "args": ["/path/to/skillslap/mcp-server/dist/index.js"],
      "env": {
        "SKILLSLAP_ENV": "local",
        "SKILLSLAP_SERVICE_ROLE_KEY": "eyJ..."
      }
    }
  }
}

With this config you can call verify_all_unverified from your agent to seed verifications on all local skills — no SQL migrations needed.

Usage Example

You: Search for testing skills
Agent: [calls search_skills with query "testing"]

You: Install the first one
Agent: [calls install_skill with the skill ID]

You: Add that skill to my agent's library
Agent: [calls slap_skill with the skill ID]

You: What skills does this agent have bookmarked?
Agent: [calls list_agent_slaps]

You: Verify all unverified skills on local
Agent: [calls verify_all_unverified with dry_run: true to preview, then without]

Changelog

v1.10.0

• Webhook tools — list_webhooks, create_webhook; manage your webhook endpoints directly from your agent
• New webhook events: skill_slapped, skill_installed, agent_session_started

v1.9.0

• Skill collections — list_collections, create_collection, add_to_collection; organize skills into named sets and compose them into CLAUDE.md bundles

v1.8.0

• Skill generation queue tools: list_skill_queue, claim_queue_item, complete_queue_item
• Agents can now process user-queued skill generation requests end-to-end

v1.7.0

• verify_skill and verify_all_unverified tools for triggering the server-side verification pipeline
• SKILLSLAP_SERVICE_ROLE_KEY support for local/dev admin operations

v1.6.0

• Skill queue for deferred generation: list, claim, and complete queue items
• generate_skill tool for AI-powered skill creation from a prompt

v1.5.0

• Agent tracking — persistent device fingerprint (~/.config/skillslap/device_id) sent on every API call; connected agents visible in dashboard at /dashboard/agents
• Agent skill library — new tools slap_skill, unslap_skill, list_agent_slaps; each agent maintains its own independent bookmark library separate from human slap counts
• New env vars: SKILLSLAP_AGENT_NAME, SKILLSLAP_MODEL

v1.4.0

• Slug-based skill URLs (@username/slug)
• username filter in search_skills
• CLI: skillslap install username/slug smart positional arg

v1.3.0

• list_my_slaps tool
• Bearer token support on all authenticated endpoints

v1.2.0

• invocation_type filter, focus_pool sort
• update_skill tool
• Unauthenticated compose_skills for public skills
• CLI list and publish commands

License

MIT