claudeception_skill

Overview

This skill automates continuous learning by extracting reusable knowledge from work sessions and codifying it into new Claude Code skills. It detects non-obvious solutions, error resolutions, and workflow optimizations, then creates or updates skills so the system improves over time. Use it to turn ad-hoc discoveries into reliable, searchable guidance.

How this skill works

Claudeception monitors sessions and evaluates whether recent work contains reusable insights. When triggered manually or after non-trivial debugging or trial-and-error, it analyzes the problem, verifies solutions, checks for existing skills, and either updates an existing skill or creates a new one with clear trigger conditions, steps, and verification. It records versioning, cross-references related skills, and includes references when web research was performed.

When to use it

• After solving a problem that required investigation or non-obvious debugging (>10 minutes)
• When a workaround, project-specific configuration, or integration trick is discovered
• If an error message was misleading and the real root cause was identified
• When a multi-step workflow can be simplified or standardized
• On explicit commands: "/claudeception", "save this as a skill", or "what did we learn?"

Best practices

• Be selective: only extract knowledge that is reusable, non-trivial, and verified
• Search existing skills first and update instead of duplicating when appropriate
• Document exact trigger conditions (error text, file names, environment) for discovery
• Include verification steps and examples so others can reproduce the fix
• Version skills: patch for wording, minor for new scenarios, major for breaking changes

Example use cases

• Discovering a Next.js server-side debugging pattern and saving it as a skill
• Extracting a workaround for a library bug that required a specific config tweak
• Saving a multi-step deployment optimization discovered during a release
• Recording the root cause and fix for a misleading CI error message
• Consolidating repeated ad-hoc fixes into a single reusable troubleshooting skill