agent-delegation
Multi-agent orchestration where a router agent delegates research questions to a research-specialist agent through the built-in `delegate` tool under default-deny delegation permissions.
Examples
Runnable starting points with README-backed workflows. Clone the skelm repo and try them locally under examples/.
agentmemory
Cross-session memory for agent steps via the agentmemory microservice. Shows observe/search/session permissions and the two-run recall pattern.
approval-workflow
Human-in-the-loop expense approval using wait() and branch(): small expenses auto-approve, larger ones pause until a manager resumes the run.
chatui-assistant
Persistent workflow exposed through a terminal or browser chat UI: one durable conversation per session, optional agentmemory, file-based persona, and an operator-gated unrestricted bypass.
dashboard-demo
Single-file HTML dashboard for the gateway /v1/dashboard/* endpoints with workflow controls, batch runs, config display, and recent error views.
incident-response
Webhook-triggered incident triage: severity branch, parallel GitHub/Slack simulation, agent root-cause analysis, ticket creation, and notification.
matrix-assistant
Persistent workflow over Matrix with per-room session state, agentmemory recall, allowed-room safeguards, and operator-gated unrestricted access.
multi-step-pipeline
Workflow mixing deterministic code, inference, a placeholder agent handoff, and final reporting to show the basic step surface end to end.
sprint-planning
Weekly cron-triggered sprint planning that fetches a simulated Jira backlog, calculates capacity, uses an agent to select stories, then notifies the team.
telegram-assistant
Persistent workflow over Telegram: a code preamble feeds a per-chat session-keyed terminal agent, with agentmemory and a file-based persona.
telegram-bot
Queue-triggered Telegram bot. The gateway owns long polling, dispatches each text update into a workflow run, and posts the result through the trigger source.
vercel-ai-bot
Minimal agent() pipeline that runs against any Vercel AI SDK model under skelm permission policy, with local OpenAI-compatible server support.