skills.rest
SmallAi-API
SmallAi-APICOMMUNITY
Explore all skills by this author →
WebsiteGitHub

agent-runtime-hooks

Registers lifecycle hooks to observe and intercept agent execution steps.

Updated Jan 27, 2026
Search Tags:#typescript#lifecycle#context compression#sub-agent#agent hooks#tool mocking#human intervention
One-click install
npx skills add https://github.com/SmallAi-API/smaihub --skill agent-runtime-hooks-smallai-api
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-runtime-hooks
Source: https://github.com/SmallAi-API/smaihub/tree/main/.agents/skills/agent-runtime-hooks
Command: npx skills add https://github.com/SmallAi-API/smaihub --skill agent-runtime-hooks-smallai-api

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Developers need visibility and control over agent execution—observing steps, mocking tool calls, handling human approvals, compressing context, and tracking sub-agent runs—without modifying core runtime code. ## Core Features & Use Cases - 16 Hook Types Across 5 Categories: Step-level, tool-call, human intervention, context compression, and sub-agent hooks registered per operation via execAgent({ hooks }). - Tool Call Mocking: beforeToolCall supports event.mock() to return fake tool results, enabling deterministic testing of agent behavior. - Human Intervention Tracking: Hooks fire before/after approval flows and when users reject and halt an agent. - Use Case: An evaluation harness (like devtools/agent-evals) uses afterStep, onComplete, afterToolCall, and beforeToolCall mocks to run reproducible agent evaluations. ## Quick Start Register an afterStep hook via execAgent to log each agent step's content, tool calls, and total cost.

Frequently Asked Questions about agent-runtime-hooks

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I mock a tool call in an agent runtime?▼

Register a beforeToolCall hook and call event.mock() with the fake result, such as event.mock({ content: '{"error":"rate limited"}' }). The dispatcher returns the mock instead of executing the real tool, but this only works in local in-memory mode.

How do I observe agent execution steps in TypeScript?▼

Pass hooks to execAgent({ hooks }) with types like beforeStep, afterStep, and onComplete. Each handler receives an AgentHookEvent containing content, toolsCalling, totalCost, and completion reason.

Does beforeToolCall mocking work with webhook hooks?▼

No, tool call mocking only works in local mode where hooks are held in memory. Webhook mode does not support mocking because dispatchBeforeToolCall requires a synchronous return value.

Why are my sub-agent hooks not firing?▼

CallAgent hooks require parentOperationId in ExecSubAgentTaskParams and are dispatched on the parent operation, not the sub-agent. Verify the parent operation ID is passed when calling execSubAgentTask.

What happens if a hook handler throws an error?▼

Hook handlers are fire-and-forget and errors are non-fatal, so a failing handler does not interrupt agent execution. The exception is beforeToolCall, which returns a mock result via dispatchBeforeToolCall.

Related Skills

Explore More Software Engineering Skills→
OpenSourceAGIOpenSourceAGI
74

agent-runtime-hooks

Registers lifecycle hooks to observe and intercept agent execution steps and tool calls.

Official
Advanced
MohamedAbdallah-14MohamedAbdallah-14
3

wz:compliance-measurement

Measure pipeline compliance for completed Wazir runs and produce per-phase and aggregate scores.

Community
Advanced
petbrainspetbrains
13

backend-pino

Configure structured JSON logging for Node.js backends with redaction and request tracing.

Official
Intermediate
mDevsLabsmDevsLabs
1

agent-runtime-hooks

Register lifecycle hooks to observe and intercept agent execution phases.

Community
Advanced