Define lifecycle hooks and decision semantics for Meerkat's agent runtime.
npx skills add https://github.com/lukacf/meerkat --skill hook-authoring-lukacf
Please help me install this Agent Skill. Skill: Hook Authoring Source: https://github.com/lukacf/meerkat/tree/main/meerkat-hooks/skills/hook-authoring Command: npx skills add https://github.com/lukacf/meerkat --skill hook-authoring-lukacf
Hook authoring for Meerkat enables developers to implement and customize lifecycle hooks, execution modes, and decision semantics that govern agent behavior.
Create a sample SKILL.md that defines hooks for RunStarted and PreToolExecution, including a patch-based rewrite example.
High-intent search queries and answers about installing and using this skill.
Lifecycle hook points like RunStarted, PreToolExecution, and RunCompleted allow you to intercept agent runtime events at specific execution boundaries. They enable custom logic injection for monitoring, safety checks, or argument modifications across Foreground, Background, and Observe execution modes.
Yes, you can rewrite tool arguments using JSON patches within the PreToolExecution hook. This patch-based rewrite mechanism allows you to dynamically alter tool inputs right before they execute, ensuring parameters meet your custom safety policies or operational requirements.
When a lifecycle hook fails, the agent runtime applies either a fail-closed or fail-open safety policy. Enforcing a fail-closed policy halts execution to prevent unsafe actions, while fail-open allows the agent to proceed despite the hook failure.
Run monitoring hooks in Observe or Background execution modes to avoid blocking the main agent runtime process. Foreground mode pauses execution until the hook completes, whereas Observe and Background modes allow non-intrusive monitoring and asynchronous task processing.
Hook authoring requires no external dependencies, but you must create a SKILL.md file with YAML frontmatter containing name and description fields. This file defines your custom hooks and decision semantics for the Meerkat agent runtime environment.
thevibeworksRead per-project plugin configuration from `.claude/plugin-name.local.md` files with YAML frontmatter.
timgranlundmarsdenDesign and validate Claude Code plugin hooks for event-driven automation.
tettuanSpawn sub-agents to execute Climpt commands via the Claude Agent SDK.
compozyConfigure MCP server connectivity and session lifecycle hooks via YAML frontmatter.
