SKILLS.md is one of the structured markdown files defining an 📝OpenClaw agent — the file that registers an agent's installable capability modules, each packaged with scripts, references, and a SKILL.md interface.
Where TOOLS.md inventories first-party APIs and integrations, SKILLS.md catalogs self-contained capability bundles. A skill might wrap a Python security-scan suite, a query-knowledge script with clear modes and flags, or any other reusable workflow the agent can call without needing to understand the internals. Each skill ships with its own SKILL.md describing the interface, parameters, and expected outputs. The agent reads SKILLS.md at boot via 📝BOOTSTRAP.md, discovers which skills are installed, and routes matching requests to the correct module.
The skills pattern is what makes OpenClaw agents extensible without modifying the agent itself. Adding a new capability means dropping a directory with a SKILL.md and supporting scripts into the agent's skills folder — no system-prompt edits, no code changes to the orchestrator. The file format mirrors the broader agentskills.io open standard, which means skills written for 📝Claude Code or other markdown-agent runtimes can be ported into OpenClaw with minimal translation. This makes the agent's capability surface as version-controllable as the rest of its identity.
Related
- 📝OpenClaw — the open-source AI assistant platform SKILLS.md belongs to
- 📝Markdown-File Agent Definition — the design pattern SKILLS.md belongs to
- 📝TOOLS.md — the parallel inventory of individual capabilities and APIs
- 📝TOOLS.md vs SKILLS.md vs MCP — clarifier disambiguating these three concepts
- 📝AGENTS.md — registry of sub-agents (workflows belong here; colleagues belong in AGENTS.md)