Agent Skills
Install the setto-api Agent Skill so Claude Code, Cursor or any AGENTS.md-aware agent knows the six SETTO External API endpoints, their populate rules and their error codes.
An Agent Skill is a folder with a SKILL.md inside:
YAML frontmatter that says what the skill is for, and a Markdown body the agent
loads when a task matches. SETTO publishes one, setto-api, that teaches an
agent the whole External API — the six endpoints, the populate allow-lists,
the error codes and the "never print the token" rule — without you explaining it
each time.
It is the fallback for every client the MCP server cannot reach, and it is useful alongside MCP too: the skill is what tells an agent when the SETTO tools are the right move.
What is in it
The skill is served from this site under /.well-known/skills/, with
/skills/… as a shorter alias for the same files.
| File | What it is |
|---|---|
setto-api/SKILL.md | The skill itself: when to use it, auth, the six endpoints with a curl each, populate, pagination, errors, safety |
setto-api/references/endpoints.md | Every parameter and every legal populate value, per endpoint, plus the data model |
setto-api/references/errors.md | Every error code, its cause, and whether a retry helps |
setto-api/cursor/setto-api.mdc | The same material as a Cursor project rule |
setto-api/AGENTS.md | A paste-in section for an existing AGENTS.md |
index.json | Machine-readable catalogue of the skills this site publishes |
The frontmatter, if you want to see the shape before installing:
---
name: setto-api
description: "Read SETTO tournament data (tournaments, divisions/categories, teams, rounds, games, standings) for one organization through the SETTO External API with an org API token. …"
license: MIT
metadata:
version: "1.0.0"
docs: "https://docs.setto.io"
openapi: "https://docs.setto.io/openapi.json"
mcp: "https://docs.setto.io/mcp"
---Install it
The skills CLI accepts a direct
download URL to a single SKILL.md, which is what this site serves:
npx skills add https://docs.setto.io/.well-known/skills/setto-api/SKILL.mdBy default it installs into the current project. Useful flags:
| Flag | Effect |
|---|---|
-g, --global | Install to your user directory instead of the project |
-a, --agent <agents...> | Target specific agents, e.g. claude-code, codex |
The CLI's other supported sources are a GitHub shorthand (owner/repo), a full
GitHub, GitLab or Azure Repos URL, any git URL, a local path, and compressed
archives. The SETTO repository is private, so the direct SKILL.md URL above is
the one to use.
Reference files
Installing from a single SKILL.md URL brings that file only. Add the two
reference files by hand if you want them — see the manual tab.
Keep the token out of it
The skill never contains a token, and it tells the agent to read one from the
SETTO_API_TOKEN environment variable and never print, log or commit it. Keep
it that way: do not paste your token into SKILL.md, setto-api.mdc or
AGENTS.md — those files get committed.
Updating
There is no auto-update. Re-run the install command to pull the current version;
the file overwrites cleanly. metadata.version in the frontmatter tells you what
you have — compare it against
SKILL.md on this site.
Breaking changes to the API are listed in the changelog.
Discovering it programmatically
https://docs.setto.io/.well-known/skills/index.json lists what this site
publishes, so a tool can find the skill without scraping the docs:
{
"skills": [
{
"name": "setto-api",
"description": "Read SETTO tournament data …",
"files": [
"SKILL.md",
"references/endpoints.md",
"references/errors.md",
"cursor/setto-api.mdc",
"AGENTS.md"
]
}
]
}Each files entry is relative to /.well-known/skills/<name>/.
Related
- MCP server — real tool calls instead of hand-written HTTP, where the client supports it.
- Docs for LLMs —
llms.txt, per-page Markdown,/openapi.json. - Authentication — where the token in
SETTO_API_TOKENcomes from.