errors_skill

This skill analyzes error patterns from Claude Code sessions to deliver actionable fixes and root-cause insights.

TypeScript

75

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill yonatangross/orchestkit --skill errors

SKILL.md2.9 KB

Overview

This skill performs error pattern analysis and troubleshooting for Claude Code sessions, surfacing recurring failures and actionable fixes. It captures runtime errors, learns common signatures, and produces reports or real-time warnings to speed diagnosis and remediation.

How this skill works

An error-collector hook records tool name, truncated error message, input that triggered the failure, timestamp, and session ID into a JSONL log. Batch analysis scripts scan recent logs to cluster repeated patterns, auto-generate rules when patterns repeat, and emit human-friendly fix suggestions or warnings for risky commands. A real-time debug endpoint inspects the current session state and flags matching rules immediately.

When to use it

After noticing recurring failures across runs to identify systemic issues
During incident response to quickly find likely root causes and fixes
Before running risky shell or database commands to get warnings
When generating a postmortem or RCA for a single bug
To monitor error trends over the last 24 hours or 7 days

Best practices

Collect errors continuously and centralize logs in the errors.jsonl file for accurate pattern detection
Run batch analysis regularly (daily/weekly) to keep auto-generated rules current
Review auto-generated rules and add manual fixes for edge cases to improve future suggestions
Use the real-time /debug flow for interactive troubleshooting of a live session
Keep fix suggestions concise and reproducible (commands, connection hints, config keys)

Example use cases

Run a 24-hour batch to surface the top PostgreSQL connection errors and get immediate remediation steps
Attach the real-time debugger to a stuck Claude Code session to detect which tool is failing
Auto-generate an error rule after two or more identical failures to trigger future warnings
Produce a markdown report summarizing error counts, signatures, and recommended fixes for an incident review
Warn users when a risky bash command matches a known dangerous pattern before execution

FAQ

Patterns are identified by clustering repeated error signatures in the JSONL logs; scripts mark a pattern when it repeats two or more times and optionally add a fix suggestion.

Where are errors stored and how do I inspect them?

Errors are written to a centralized JSONL log (errors.jsonl) with tool, truncated message, input, timestamp, and session ID; you can query this file or run the batch analyzer to generate reports.