Execute arbitrary code before or after AI agent lifecycle events.
npx skills add https://github.com/xkl2013/lobe --skill agent-runtime-hooks-xkl2013
Please help me install this Agent Skill. Skill: agent-runtime-hooks Source: https://github.com/xkl2013/lobe/tree/main/.agents/skills/agent-runtime-hooks Command: npx skills add https://github.com/xkl2013/lobe --skill agent-runtime-hooks-xkl2013
This Skill provides fine-grained control over the lifecycle of AI agents, allowing users to hook into key points during execution to perform various operations like tool execution, human intervention, context compression, sub-agent calls, and more.
To create a custom hook, define the hook type and handler in the SKILL.md file and then use execAgent to register it.
High-intent search queries and answers about installing and using this skill.
AI agent lifecycle hooks let you execute arbitrary code at key execution points by defining hook types and handlers in SKILL.md and registering them via execAgent, covering beforeStep, afterStep, and tool call boundaries.
Yes, you can catch errors during tool execution and human intervention by registering onToolCallError, onCallAgentError, and onCompactError hooks to handle exceptions at specific lifecycle events.
Available events for context compression and sub-agent calls include beforeCompact, afterCompact, onCompactError, beforeCallAgent, afterCallAgent, and onCallAgentError hooks to control execution flow.
You can track and log each step of an AI agent's execution flow by defining a beforeStep hook to trigger logging operations before each step is executed during the agent lifecycle.
No specific dependencies are required to set up human intervention hooks. You can register beforeHumanIntervention, afterHumanIntervention, and onStopByHumanIntervention handlers directly within your agent environment.
compozyDetect invalid AGH hook commands during session.post_create registration.
mDevsLabsRegister lifecycle hooks to observe and intercept agent execution phases.
zmh13928433602-sysRegister lifecycle hooks to observe and intercept agent execution events.
duwenjiIntercept agent lifecycle events and mock tool calls via hooks.
