How does this skill handle API rate limits?

It implements Tier 1 recovery using exponential backoff with jitter, which waits for increasing intervals before retrying to ensure the service becomes available without overloading it.

Can it help fix broken builds or tests?

Yes, it includes logic to parse error messages from build tools or test runners, identify the failing component, and attempt automated fixes for common syntax or dependency issues.

What is cascading error recovery?

Cascading recovery is a structured approach that attempts simple fixes like retries first, then moves to more complex solutions like rephrasing prompts or switching models before finally involving a human.

When does the skill escalate to a human?

Escalation occurs at Tier 4 after all automated recovery attempts (retries, fallbacks, and substitutions) have failed, or when a blocking issue like an authentication failure is detected.

Error Recovery for Claude Code

Name: Error Recovery for Claude Code
Author: pagerguild

bypagerguild

0•

Productivity & Workflow

Implements cascading recovery strategies to handle tool failures, rate limits, and multi-agent orchestration errors automatically.

The Error Recovery skill provides a structured framework for managing failures in complex multi-agent workflows within Claude Code. By implementing a four-tier cascading recovery system—ranging from transient error retries with exponential backoff to semantic fallbacks, agent substitution, and structured human escalation—it ensures that automated tasks are resilient to network issues, model hallucinations, and context window limitations. This skill is essential for developers building robust AI-driven automation where reliability and graceful degradation are critical for maintaining continuous development cycles.

Key Features

01Exponential backoff with jitter for transient API and rate limit errors

02Four-tier cascading recovery strategy for tiered error handling

03Circuit breaker pattern to prevent cascading failures in complex workflows

040 GitHub stars

05Built-in observability for tracking error rates and recovery success metrics

06Semantic fallback prompts to correct invalid JSON or hallucinated outputs

Use Cases

01Routing complex tasks to more capable models when context window limits are reached

02Automatically fixing build or test failures by diagnosing error logs and logs

03Managing API rate limits and network timeouts during long-running agent tasks

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add pagerguild/guilde-lite error-recovery

For use in Claude.ai and ChatGPT

Download Skill