How are rate limits (429 errors) managed?

It automatically detects rate limiting, respects 'Retry-After' headers if provided, and implements exponential backoff to retry requests before eventually falling back.

How does this skill handle model timeouts?

It monitors execution time against predefined thresholds and provides strategies to retry with increased limits, skip the stalled model, or notify the user for manual intervention.

What happens if an external API key is invalid?

The skill detects 401 Unauthorized errors and triggers a fallback mechanism that switches to the embedded Claude model while providing a clear error message to the user.

Does it support partial success in multi-model reviews?

Yes, it includes logic to proceed with consensus-building as long as a minimum threshold of models (e.g., 2 out of 4) respond successfully, ensuring the workflow doesn't fail entirely.

Error Recovery & Resilience

Name: Error Recovery & Resilience
Author: MadAppGang

byMadAppGang

•

177

•

Productivity & Workflow

Implements robust error-handling patterns for multi-agent workflows to ensure graceful recovery from timeouts, API failures, and partial successes.

The Error Recovery skill provides battle-tested patterns for building resilient AI workflows that can survive the inherent instability of external dependencies. It enables developers to handle model timeouts, 4xx/5xx API errors, and rate limits using strategies like exponential backoff and graceful degradation. By implementing partial success logic, your multi-agent systems can continue delivering results even when individual components fail, ensuring that your automation remains production-ready and reliable under any network conditions.

Key Features

01177 GitHub stars

02Automated timeout detection and mitigation for slow AI model responses

03Intelligent fallback to embedded models when external services are unreachable

04Comprehensive API failure recovery for 401, 429, and 500 status codes

05Partial success strategies to maintain consensus in multi-model workflows

06Graceful cleanup and state saving for user cancellations (Ctrl+C)

Use Cases

01Developing resilient CI/CD integrations that require high availability from external models

02Building production-grade multi-agent code review systems that survive API blips

03Automating complex, long-running data processing tasks across multiple LLMs

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add madappgang/claude-code error-recovery

For use in Claude.ai and ChatGPT

Download Skill