skills.rest
seiggy
seiggyCOMMUNITY
Explore all skills by this author →
GitHub

error-recovery

Applies retry, fallback, diagnosis, escalation, and degradation patterns to agent failures.

1|Updated Jul 7, 2026
Search Tags:#error recovery#escalation#fallback#graceful degradation#agent reliability#failure handling#retry backoff
One-click install
npx skills add https://github.com/seiggy/maf-copilot-studio-demo --skill error-recovery-seiggy
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: error-recovery
Source: https://github.com/seiggy/maf-copilot-studio-demo/tree/main/.copilot/skills/error-recovery
Command: npx skills add https://github.com/seiggy/maf-copilot-studio-demo --skill error-recovery-seiggy

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agents often stop at the first failure and simply report the error instead of recovering. This Skill gives agents a standard playbook for handling transient errors, broken tools, build failures, and partial outages so tasks keep moving forward. ## Core Features & Use Cases - Five Recovery Patterns: Retry with Backoff, Fallback Alternatives, Diagnose-and-Fix, Escalate with Context, and Graceful Degradation, each with clear triggers and step-by-step procedures. - Failure-Type Selection Guide: A mapping table that pairs common failure types (network, auth, build, missing data) with the right primary and fallback pattern. - Use Case: An agent's API call returns 429 Too Many Requests. Instead of failing, it waits 2s, retries, doubles the wait each attempt, and escalates with full context only after three failed retries. ## Quick Start Apply the error-recovery patterns to handle any failure you encounter during this task instead of stopping at the first error.

Frequently Asked Questions about error-recovery

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

FAQPage Schema
How do I make an AI agent retry failed API calls?▼

Use the Retry with Backoff pattern: wait 2 seconds, retry, and double the wait each attempt up to 3 retries. Log each attempt with the error received, then escalate with full context if the call still fails.

What should an agent do when a build or test fails?▼

Use the Diagnose-and-Fix pattern: read the full error output, identify the root cause, apply a targeted fix, and re-run to verify. Allow a maximum of 3 fix-retry cycles before escalating to a human or coordinator.

When should an agent escalate an error instead of retrying?▼

Escalate when recovery attempts are exhausted or the failure requires human judgment, such as unknown or novel errors. Include what was attempted, exact error messages, the suspected root cause, and suggested next steps.

Can an agent continue a task when part of it fails?▼

Yes, using Graceful Degradation. If the failed step is non-critical, log the failure, deliver partial results with a clear note of what was skipped, and offer to retry the skipped step separately.

What are the limitations of automated error recovery patterns?▼

These patterns do not fix design flaws or novel errors with no known cause. Retry is capped at 3 attempts, and unknown errors route directly to escalation rather than automated recovery.

Related Skills

Explore More Software Engineering Skills→
wedsamuel1230wedsamuel1230
1

dispatching-parallel-agents

Dispatch independent agents in parallel to investigate distinct failure domains.

Community
Basic
blakeoxblakeox

postmortem

Reconstruct incident timelines and perform root cause analysis from logs.

Community
Advanced
MonumentalSystemsMonumentalSystems
17

monitoring-observability

Establish monitoring and observability across metrics, logs, and traces.

Official
Advanced
apankov1apankov1
5

fault-injection-testing

Inject deterministic storage and network faults to test circuit breaker and retry resilience.

Community
Advanced