Skip to content

index.md

Latest commit

 

History

History
111 lines (86 loc) · 3.98 KB

index.md

File metadata and controls

111 lines (86 loc) · 3.98 KB

Reference

Public API, grouped by area. Every name listed here is importable directly from the top-level skillengine package unless noted.

Core

Symbol Purpose
SkillsEngine Loads, filters, snapshots, and executes skills
SkillsConfig Engine configuration (skill dirs, per-skill overrides)
SkillEntryConfig Per-skill enable/disable/override knobs
CacheRetention Snapshot cache policy
Skill, SkillMetadata, SkillRequirements Core dataclasses
SkillSnapshot Immutable view returned by engine.get_snapshot()

Agent

Symbol Purpose
AgentRunner Conversation orchestrator
AgentConfig Model, base URL, API key, max turns, tool flags
AgentMessage A single message in the conversation log
TextContent, ImageContent Multi-modal message parts

Key methods: chat, chat_structured, chat_stream_events, connect_mcp_servers, disconnect_mcp_servers, add_handoffs, remove_handoffs, switch_model.

Events

EventBus with handler registration via on(event_type, handler, priority=0). Event types live in skillengine.events.

BEFORE_TOOL_CALL handlers can return ToolCallEventResult(block=True) to cancel a tool call.

Tools

ToolDispatcher registers ToolDefinition objects via dispatcher.register(name, handler, definition) where handler has the shape async (name, args, ctx, on_output) -> str and definition is an OpenAI function-calling object.

Built-in tools: BashTool, ReadTool, WriteTool, EditTool, GrepTool, FindTool, LsTool.

Adapters

Symbol Purpose
LLMAdapter Abstract base
OpenAIAdapter OpenAI + OpenAI-compatible
AnthropicAdapter Claude with adaptive thinking
AdapterRegistry, AdapterFactory Lazy instantiation

MCP

Symbol Purpose
MCPClient JSON-RPC stdio client (tools/list, tools/call, …)
MCPServer Server side, exposes skills as skill__<name>
MCPServerSpec Transport, command, args, env, cwd
MCPConnectionPool Manages multiple MCPClients + dispatcher integration
parse_mcp_uri(uri) Parse mcp+stdio:…, mcp+stdio://…, mcp+command:{…}

A2A / Handoffs

Symbol Purpose
AgentCard, AgentCardSkill, AgentCapabilities A2A protocol cards
AgentRegistry, RegisteredAgent, AgentSource, AgentStats Registry
A2AClient Send tasks to remote A2A agents
AgentDiscovery, DiscoveryConfig, AgentHealthStatus Auto-discovery
PerformanceRouter, RoutingConfig, RouteResult Latency-aware routing
CoordinatorAgent, CoordinatorConfig Multi-agent orchestrator
Handoff, handoff, callable_handoff, agent_handoff, a2a_handoff OpenAI Agents / Anthropic A2A compat shim

Structured output

Symbol Purpose
extract_json_schema(T) JSON schema for T (Pydantic / dataclass / TypedDict / dict)
extract_json_payload(text) Pull a JSON value out of arbitrary text
parse_structured(T, text) Combine extract + validate
build_directive(T) Prompt suffix appended by chat_structured
StructuredOutputError Raised on unrecoverable parse / validation failures

CLI

uv run skills list -d ./skills        # list eligible skills
uv run skills show <name>             # print a skill's body
uv run skills validate -d ./skills    # validate names/descriptions
uv run skills prompt -d ./skills      # print the system prompt
uv run skills exec <name> ...         # execute a skill's commands
uv run skills watch -d ./skills       # hot-reload

python -m skillengine.mcp --skill-dir ./skills [--name N] [--version V]
                                              [--instructions S]

Runtimes

Symbol Purpose
BashRuntime Shell execution with timeout + streaming
CodeModeRuntime search + execute pattern over an arbitrary spec
BoxLiteRuntime (optional) VM-level isolation via micro-VMs