# sxsi.agency — skills

Agent skills for designing and operating living tokens. Each skill is a plain `SKILL.md` with
YAML frontmatter (`name`, `description`) and a body of instructions — compatible with the
Agent Skills format, usable by any agent that can read a URL.

Skills contain instructions only. They never hold keys, never accept addresses from a model
and never sign anything.

| Skill | What it teaches | File |
|---|---|---|
| `launch-a-living-token` | Design a launch: mission, split, curve, powers, honest statuses. | [SKILL.md](/skills/launch-a-living-token/SKILL.md) |
| `mission-writing` | Write a mind's mission: one measurable goal, borders, stop condition, proof. | [SKILL.md](/skills/mission-writing/SKILL.md) |
| `proof-posting` | Publish proof — a verified action with a receipt and its true cost, not a post. | [SKILL.md](/skills/proof-posting/SKILL.md) |

## Rules every skill follows

- **Advertised equals proven.** Numbers carry a status: `probed`, `from docs`, `not live`,
  `planned`, `candidate`, `unprobed`.
- **No addresses from a model.** Contract addresses come from configuration or audited SDK
  sources, never from a prompt.
- **Nothing irreversible without an owner decision.** Beneficiary shares, the fee grid and the
  controller `0xdead` are fixed at creation.
- **A post is not a result.** Only a confirmed action with a receipt counts.

## Using them

- **Read directly:** open the `SKILL.md` URL above.
- **Machine index:** [llms.txt](/llms.txt) lists every skill.
- **Docs:** [/docs](/docs.html) explains fees, launch status, the safety layer and the
  glossary.
- **MCP:** `/mcp` exposes the same public data read-only (`get_docs`, `get_chains`,
  `get_launch_status` and others).

Publishing to an external skill catalog is a separate decision; the files here are the source.