AgentSkillsCN

agent-operations

管理MCPBundles自主智能体。在创建智能体、审查运行历史，或调试智能体故障时使用。

SKILL.md
--- frontmatter
name: agent-operations
description: Manage MCPBundles autonomous agents. Use when creating agents,
  reviewing run history, or debugging agent failures.

Agent Operations

Create

upsert_agent(name, agents_md_content, heartbeat_md_content, mcp_source={"type":"bundle","bundle_slug":"..."}, ai_provider, model_name, heartbeat_interval_minutes, is_active, max_tool_calls_per_run)

Skills

upsert_agent_skill(agent_id, skill_name, skills_md_content) — add composable instructions to an agent's prompt. delete_agent_skill(id) to remove.

Monitor

  • •list_agents — all agents with status and last heartbeat.
  • •list_agent_runs(agent_id="...") — run history with status, duration, tool calls, tokens, errors.
  • •trigger_agent_run(agent_id="...") — manual run, returns run ID.

Debug failures

  1. •list_agent_runs(agent_id="...", status="error", limit=5) — find failed runs.
  2. •list_agent_runs(id="...") — full run details and error output.
  3. •check_readiness(bundle_slug="...") — is the MCP source bundle ready?
  4. •list_executions(heartbeat_run_id="...") — tool executions from that run.

Lifecycle

Pause: upsert_agent(id, is_active=false). Resume: is_active=true. Delete: delete_agent(id).