tool-use-guardian

Automate tool-call retries and recovery for multi-step AI agent workflows.

27|5|Updated Apr 3, 2026
One-click install
npx skills add https://github.com/Fandry96/k3-agentic-skills --skill tool-use-guardian
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: tool-use-guardian
Source: https://github.com/Fandry96/k3-agentic-skills/tree/main/skills/tool-use-guardian
Command: npx skills add https://github.com/Fandry96/k3-agentic-skills --skill tool-use-guardian

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Reliable tool orchestration for AI agents by automatically detecting tool-call failures, retrying operations, and learning from past errors to prevent cascading issues.

Core Features & Use Cases

  • Pre-call validation ensures required parameters exist and are correctly typed before invoking any tool.
  • Automatic retries with configurable limits and fallback strategies for flaky tools.
  • Failure classification and checkpoint-based recovery to resume long chains after errors.
  • Learning from failure patterns to mark tools as unreliable and suggest alternatives.
  • Use Case: An agent performing a multi-step workflow with web requests, database calls, and API calls, where transient failures previously halted progress.

Quick Start

Wrap your tool calls with Guardian to automatically validate inputs and recover from failures.

Frequently Asked Questions about tool-use-guardian

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

FAQPage Schema
How do I handle AI agent tool-call failures and auto-retry timeouts or rate limits?

Handle AI agent tool-call failures by wrapping calls with Guardian to auto-retry timeouts and rate limits. It detects failures and applies recovery strategies to prevent cascading issues across multi-step workflows without code changes.

What is the best way to recover from truncated JSON in multi-step tool workflows?

The best way to recover from truncated JSON in multi-step tool workflows is using checkpoint-based recovery. Guardian detects mid-chain errors, classifies failures, and resumes long chains from the last successful checkpoint.

How do I validate tool call parameters before invoking external APIs in an agent workflow?

Validate tool call parameters before invoking external APIs by applying pre-call validation. Guardian checks that required parameters exist and match correct types before any tool executes, preventing malformed requests.

Can I automatically mark unreliable tools and suggest alternatives after repeated agent failures?

Yes, you can automatically mark unreliable tools and suggest alternatives. Guardian learns from failure patterns across tool calls, identifying flaky tools and applying fallback strategies to improve reliability.

Does tool-use-guardian require code changes to add guardrails for flaky API calls?

No, tool-use-guardian does not require code changes to add guardrails for flaky API calls. You wrap existing tool calls with Guardian to automatically apply validation, retry limits, and fallback strategies.

Why does my long chain of tool calls halt progress after a transient web request failure?

Long chains of tool calls halt progress after transient web request failures because standard execution lacks auto-recovery. Guardian fixes this by classifying failures and using checkpoint-based recovery to resume workflows.