Skills

A skill is a reusable, provider-independent capability that agents can invoke.

class Skill(ABC):
    @property
    def name(self) -> str: ...
    @property
    def description(self) -> str: ...
    @property
    def parameters(self) -> dict: ...  # JSON Schema

    async def execute(self, params: dict) -> SkillResult: ...

Skills map directly to "tools" in LLM APIs but are defined once and work across all providers.

Built-in Skills

Skill	Description
file_read	Read file contents
file_write	Write content to file
glob_search	Search files by pattern
shell_exec	Execute shell commands
sandboxed_shell	Execute commands in Docker sandbox
web_read	Fetch web page content
doc_sync	Documentation sync checker
github	GitHub integration via gh CLI
webhook	Outgoing webhook skill
load_skill	On-demand full skill instruction loading
ask_clarification	Request human clarification (blocking/non-blocking)

Skills Map (Agent Assignments)

Skill	Agent	Description
/docker-build	devops	Build and manage containers via OrbStack
/test-runner	all	Run pytest suite via Docker
/lint-check	all	Ruff linting and formatting checks
/code-review	all	Automated quality/security review
/deploy	devops	Container deployment via docker-compose
/scout	scout	GitHub pattern discovery
/website-dev	frontend	Documentation site development
/verify	all	Pre-PR quality gate (tests, lint, format, security, diff review)
/cost-optimization	ai-engineer	Review LLM API costs, routing, budget, retry efficiency
/ship	all	Full pipeline: test, lint, docs sync, commit, push
/feature	all	End-to-end feature dev: implement, tests, SOLID review, docs, commit
/fix	all	Bug fix with mandatory regression tests, lint, deploy
/doc	all	Full docs review: audit all docs/ against codebase
/fetch-star-repos	scout	Fetch GitHub starred repos for research scout
/research-scout	research-scout	Analyze starred repos and propose code improvements
/web-research	all	Search the internet for solutions, docs, and best practices

Skill Registry

registry = SkillRegistry()
registry.register(FileReadSkill())
registry.register(ShellExecSkill(timeout=120))

# Export as tool definitions for any LLM
tools = registry.to_tool_definitions()