Skip to content
This repository was archived by the owner on Feb 1, 2026. It is now read-only.

Configuration

Spacehunterz edited this page Dec 11, 2025 · 1 revision

Configuration

CLAUDE.md

The ~/.claude/CLAUDE.md file contains instructions Claude follows. The framework installs a template that:

  • Requires querying the building before tasks
  • Lists the golden rules
  • Explains how to record learnings

You can add your own instructions below the ELF section.

settings.json

Hook configuration in ~/.claude/settings.json:

{
  "hooks": {
    "PreToolUse": [{
      "matcher": "Task",
      "hooks": [{
        "type": "command",
        "command": "python ~/.claude/hooks/learning-loop/pre_tool_learning.py"
      }]
    }],
    "PostToolUse": [{
      "matcher": "Task",
      "hooks": [{
        "type": "command",
        "command": "python ~/.claude/hooks/learning-loop/post_tool_learning.py"
      }]
    }]
  }
}

Key Phrases

The CLAUDE.md template teaches Claude to respond to:

Phrase What It Does
check in Queries building, reports context
query the building Same as check in
what does the building know about X Searches for topic X
record this failure: [lesson] Creates failure log
record this success: [pattern] Documents success
start experiment: [hypothesis] Tracks hypothesis

Automatic vs Manual Learning

Automatic (via hooks):

  • Pre-task query injects relevant heuristics
  • Post-task recording logs outcomes
  • Confidence updates based on success/failure
  • Trail laying tracks files touched

Manual (via key phrases):

  • Major failures worth documenting
  • Project-specific patterns
  • Experiments you want to track

Clone this wiki locally