Learn Error Pattern Skill

Learn Error Pattern Skill

Parse a user-provided "error -> solution" pair, classify it, store it in the cross-session learning database at high confidence, and confirm back. One pattern per invocation. All database operations go through the learning-db.py CLI.

Instructions

Step 1: Parse Input

Extract two fields from the user's input:

• error_pattern: The error message or symptom text
• solution: The fix or resolution text

Accepted input formats:

• /learn "error pattern" -> "solution"
• /learn "error pattern" => "solution"
• Freeform: "teach that X means Y" or "remember: when X, do Y"

Both fields must be non-empty. If either is missing, ask the user for the missing part before proceeding. If the error pattern is vague (e.g., "it broke") or the solution is non-actionable (e.g., "fix it"), ask the user to provide the specific error message and concrete fix steps — vague patterns fail to match future errors and waste database space.

Step 2: Classify Fix Type

Determine fix_type and fix_action from the solution text by applying these rules in order:

1. Solution contains an install command (pip install, npm install, apt install) -> fix_type=auto, fix_action=install_dependency
2. Solution contains replace_all -> fix_type=auto, fix_action=use_replace_all
3. Solution references a skill name -> fix_type=skill, fix_action=<skill-name>
4. Solution references an agent name -> fix_type=agent, fix_action=<agent-name>
5. Otherwise -> fix_type=manual, fix_action=apply_suggestion

Step 3: Store Pattern

Execute the learning-db.py CLI to persist the pattern. Always pass user-provided strings as CLI arguments exactly as shown — never inline them into Python code via f-strings or string concatenation, because quotes or special characters in error text will break the script and create injection risk.

python3 ~/.claude/scripts/learning-db.py record \
  "<error_type>" \
  "<error_signature>" \
  "<error_pattern> → <solution>" \
  --category error \