skelm
Meta-package — install this. Re-exports @skelm/core, ships the skelm CLI binary, and depends on the CLI, scheduler, and integration SDK pieces needed for local execution.
Packages
skelm ships as focused packages. Install the meta-package `skelm` for the runtime and CLI, then add backend, integration, memory, metrics, or tracing packages as needed.
@skelm/core
Runtime, types, builders, schemas, events, permissions, registries, backend interface, system prompt composition, and pipeline execution primitives.
@skelm/cli
Command-line interface and programmatic primitives. Non-exempt commands dispatch to the local gateway; `init`, `validate`, and `gateway *` can run without a live gateway.
@skelm/gateway
Long-running orchestrator that owns config, registries, permissions, audit, agent lifecycle, triggers, HTTP/SSE, dashboard API, and the execution trust boundary.
@skelm/scheduler
Trigger management for cron, interval, and webhook schedules with deduplication and overlap policies; used by the gateway for scheduled work.
@skelm/integrations
Typed integration package for third-party services such as GitHub, Slack, Jira, Telegram, Matrix, and chat UI trigger sources.
@skelm/integration-sdk
Authoring SDK for custom skelm integrations and trigger sources. Use it to build connectors that plug into the gateway.
@skelm/agent
First-party native agent backend. Runs infer() and agent() against OpenAI-compatible chat endpoints with in-process TrustEnforcer permission checks.
@skelm/agentmemory
Typed REST client and gateway-wired AgentmemoryHandle for cross-session recall. Operations are default-deny and routed through gateway enforcement.
@skelm/opencode
Opencode.ai coding-agent backend with native permission mapping, granular enforcement, and gateway-supervised lifecycle support.
@skelm/pi
Pi coding-agent SDK backend with native tool allowlist enforcement, infer() support, sandbox defaults, and queueing controls.
@skelm/codex
OpenAI Codex backend via @openai/codex-sdk with boundary permission mapping, MCP injection, skill loading, session lifecycle, and streaming.
@skelm/vercel-ai
Vercel AI SDK backend that powers infer() and agent() with tool filtering and call-time permission checks under skelm policy.
@skelm/metrics
Prometheus-format metrics for skelm event streams: run counters, step timings, permission denials, approvals, and trigger fires.
@skelm/otel
OpenTelemetry tracing for skelm event streams, emitting run and step spans without configuring exporters for you.