Super Agent Skill
Start

Connect · MCP

One MCP server.
Every coding agent.

Plug Super Agent Skill into Lovable, Claude, Cursor, Codex, VS Code, Zed, n8n and any MCP-compatible runtime. Read tools work anonymously — write tools need a token.

Streamable HTTP endpoint

https://superagentskill.com/api/mcp
Live·3 read tools · free·4 write tools · token

Plug and play · one command

OAuth login + client auto-config, no JSON editing

The CLI runs the full OAuth flow in your browser (loopback PKCE), stores the token locally, and writes the MCP config for your client — including the auth-gated write tools. It also ships a local stdio bridge that injects and refreshes the token automatically.

terminal
npx super-agent connect --client claude-code
# clients: claude-code · claude · cursor · codex · vscode · windsurf

npx super-agent status     # token state
npx super-agent logout     # revoke + forget

Stdio-only client? Point it at npx -y super-agent mcp — the bridge handles auth for you.

01 · Fastest path

Don't touch JSON. Paste a prompt.

Pick your agent. Paste the prompt in chat. It writes the config, restarts, and runs a test call so you know it works.

Paste in One prompt → fully connected (recommended)
Connect me to the "Super Agent Skill" MCP server with full access (read + write tools).

Run this in my shell and follow the browser login it opens:

  npx -y super-agent connect --client claude-code

Notes:
- It does OAuth (loopback PKCE), stores the token in ~/.superagentskill, and writes my client's MCP config automatically.
- Swap --client for: claude | cursor | codex | vscode | windsurf.
- After it finishes, restart the client, then call the MCP tool "overview" and show me the intent→tool map.
If anything fails, paste the exact command output and the config file it wrote.

For write tools (upload_packages, request_primitive), append Authorization: Bearer <token> from Account → Tokens.

02 · Verify

Test the MCP server, right here.

Probe the live endpoint from your browser or a shell. No setup needed.

Test MCP live

Same endpoint your client will hit

Calls POST /api/mcp with the proper Accept header. Read tools work anonymously; write tools need a token from /account/tokens.

Authorization (optional)

Sent as Authorization: Bearer …. Mint one at /account/tokens.

Or hit the public health endpoint from your terminal

Confirms reachability from your network and enumerates the live tool catalog — no auth.

shell
curl -s https://superagentskill.com/api/public/mcp/health | jq

03 · Configs

Manual setup for every client.

Pick your client. Copy the snippet. Restart. Tools appear in the agent.

Lovable

Workspace

1-click

Add Super Agent Skill as an MCP connector inside the Lovable workspace. Tools become available to the agent during builds.

  1. 1Open Connectors → Add MCP Server in your Lovable workspace.
  2. 2Paste the URL: https://superagentskill.com/api/mcp
  3. 3Save. Tools appear in the agent automatically on the next message.
Workspace → Connectors
Name: Super Agent Skill
URL:  https://superagentskill.com/api/mcp
Auth: none (read-only) · Bearer <token> for write tools

04 · Power moves

Once connected — what to ask the agent.

Three intent groups. Copy a prompt, swap a path or topic, send. The agent routes to @superagentskill automatically.

Significantly improve a local skill, playbook, soul or guardrail. Uses get_methodology → review_skill → host edits → re-score. The score delta is the proof.

Upgrade a skill end-to-end
Use @superagentskill to significantly upgrade my local skill at .claude/skills/code-reviewer.md.
1. Call get_methodology to load the rubric.
2. Call review_skill on the file's raw content — show me the per-pillar score.
3. Edit the file in place applying every top_action.
4. Re-run review_skill and show the before/after grade and per-pillar delta.
Stop only when the overall score is >= 90.
Audit & harden a playbook
Use @superagentskill to audit docs/playbooks/incident-response.md against the SuperAgentSkill methodology.
Score it, then rewrite the 3 weakest pillars in place.
Show the diff and the new overall_score. Don't touch pillars already at >= 80.
Harden a soul / persona
Use @superagentskill to upgrade the soul at agents/nova/SOUL.md.
Focus on Identity, Guardrails and Trust pillars.
After review_skill, also call search_registry type="soul" to borrow patterns from the 3 highest-trust souls. Merge what fits, then re-score.
Author a new guardrail
Use @superagentskill: call get_methodology, then draft a brand-new guardrail file at .agents/guardrails/no-pii-leak.md that scores >= 90 on every pillar. Verify by calling review_skill on your draft BEFORE handing it to me.
Batch-upgrade a folder
Use @superagentskill to upgrade every .md file under .claude/skills/.
For each file: review_skill → edit in place → re-review. Skip files already at grade A.
At the end, print a table: file | before | after | delta.
Compare two versions
Use @superagentskill to score both versions of my skill (skill-v1.md and skill-v2.md) with review_skill.
Tell me which one wins per pillar, what v2 broke that v1 had right, and what to merge into a v3.

Pro tips

  • → If the agent can't find the server, ask it to call overview first — it returns the full intent → tool map.
  • → Always run review_skill twice (before and after edits). The score delta is the proof.
  • → Always call get_skill_trust before recommending a registry primitive.
  • → If your client doesn't auto-mention MCP servers, just write: "Using the superagentskill MCP, …"

Don't see your client?

Any runtime that speaks MCP Streamable HTTP connects with the same URL. The full reference (tools, schemas, errors) lives in the MCP docs.