What Are Skills?
Skills are self-contained capability modules. Each skill ships:- A script directory with TypeScript scripts the agent can invoke
- A
SKILL.mdmanifest that tells the agent exactly what it can do, what parameters to pass, and what to expect back
SKILL.md at startup and immediately knows how to call the skill — no hardcoded capability lists in your system prompt, no outdated documentation.
The Manifest System
When you provision a wallet, aGLOSSO.md is written to your working directory. It aggregates all installed skills into one file:
Available Skills
glosso-wallet
SOL balance checks, transfers, and transaction history.
glosso-pyth
Real-time price feeds for SOL, BTC, ETH, USDC, JUP, BONK and more via Pyth Network.
glosso-jupiter
Token swap quotes and execution via the Jupiter aggregator.
OpenClaw install
Install all skills onto an OpenClaw agent VM with one command.
How Skills Are Loaded
At provision time, Glosso writesGLOSSO.md to the current directory by scanning ~/.openclaw/skills/glosso-*/SKILL.md (or the local packages/skills/ directory). The GLOSSO.md is a concatenation of each skill’s capabilities section.
In your agent:
Anatomy of a SKILL.md
EachSKILL.md is a machine-readable manifest with frontmatter and structured sections:
description field is injected into the agent’s tool discovery. The body documents each function the agent may call.
Skills are just TypeScript scripts + a manifest. You can write your own — see Writing a Skill.