Claude Code Plugins

Community-maintained marketplace

Feedback

error-recovery

@costiash/CognivAgent
4
0

Provides structured error handling protocol for all agent operations. Ensures users receive clear error reports, troubleshooting guidance, and actionable next steps. Use when any tool operation fails, encounters an error, or produces unexpected results.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name error-recovery
description Provides structured error handling protocol for all agent operations. Ensures users receive clear error reports, troubleshooting guidance, and actionable next steps. Use when any tool operation fails, encounters an error, or produces unexpected results.

Error Recovery

On Failure Protocol

  1. STOP IMMEDIATELY — Do not continue with other actions
  2. REPORT CLEARLY — Explain what went wrong concisely
  3. OFFER NEXT STEPS — Provide specific troubleshooting or retry options
  4. WAIT FOR RESPONSE — Do NOT proceed autonomously

Error Response Template

[Operation] Failed

[Brief explanation of what went wrong]

This can happen when:
- [Possible cause 1]
- [Possible cause 2]

**Would you like to:**
1. [Retry option]
2. [Alternative option]
3. Cancel and return to main menu

Please let me know how you'd like to proceed.

Common Error Patterns

Category Error Troubleshooting
Transcription YouTube URL invalid Check URL format and video availability
Transcription File not found Verify file path exists
Transcription FFmpeg missing Install FFmpeg (apt install ffmpeg)
Transcription API key missing Set OPENAI_API_KEY environment variable
Transcription Timeout Video too long; suggest splitting
KG Bootstrap failed Transcript may be too short or lack entities
KG Extraction failed Project may not be bootstrapped yet
KG Project not found Check project ID validity

NEVER Do

  • Try to work around failures without user consent
  • Make assumptions about what went wrong
  • Proceed with alternative approaches autonomously
  • Leave the user stuck without a path forward

This saves time and money by avoiding unnecessary API calls when something has already failed.