Can this skill identify unreliable external services?

Yes, the learning engine tracks failure patterns per tool and proactively marks a tool as unreliable after three or more repeated failures, suggesting alternatives to the user.

How does the checkpoint system prevent task restarts?

Guardian maintains state throughout multi-step tool chains, allowing the agent to resume from the specific step that failed rather than restarting the entire sequence from scratch.

Is Tool Use Guardian compatible with various AI models?

While optimized for Claude Code, it is designed to work with a broad spectrum of tools and models including Cursor, Gemini, Copilot, and Windsurf.

What common failures does Tool Use Guardian resolve?

It automatically handles 9 major failure types, including truncated JSON, API timeouts, rate limits (429), authentication errors, and schema mismatches.

Tool Use Guardian

Name: Tool Use Guardian
Author: JantonioFC

byJantonioFC

•

分析と監視

Protects AI agent workflows by automatically monitoring, retrying, and recovering from failed tool calls and API interruptions.

Tool Use Guardian acts as an essential reliability layer for AI agents, specifically designed to prevent task failure when interacting with external tools and APIs. By implementing intelligent retry logic, failure classification across nine distinct categories, and automatic recovery from common issues like truncated JSON or rate limits, it ensures that your agent's chain of thought remains unbroken. Its sophisticated checkpoint system allows multi-step tasks to resume exactly where they failed, while its learning engine identifies consistently unreliable tools to help you optimize your agent infrastructure.

主な機能

012 GitHub stars

02Sophisticated retry logic with exponential backoff for rate limits

03Intelligent auto-recovery from truncated JSON and malformed responses

04Multi-step chain protection with checkpoint-based task resumption

05Classification and automated handling for 9 distinct tool failure types

06Automated pattern learning to identify and flag unreliable tool behaviors

ユースケース

01Stabilizing complex agentic workflows that rely on multiple external API calls

02Preventing task interruptions during temporary network instability or rate limiting

03Ensuring data consistency when tools return incomplete or schema-mismatched data

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jantoniofc/skillsbank tool-use-guardian

For use in Claude.ai and ChatGPT

主な機能

012 GitHub stars

02Sophisticated retry logic with exponential backoff for rate limits

03Intelligent auto-recovery from truncated JSON and malformed responses

04Multi-step chain protection with checkpoint-based task resumption

05Classification and automated handling for 9 distinct tool failure types

06Automated pattern learning to identify and flag unreliable tool behaviors

ユースケース

01Stabilizing complex agentic workflows that rely on multiple external API calls

02Preventing task interruptions during temporary network instability or rate limiting

03Ensuring data consistency when tools return incomplete or schema-mismatched data

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jantoniofc/skillsbank tool-use-guardian

For use in Claude.ai and ChatGPT