Parsimony for LLMs: Knowing When It’s Good Enough

prompting-for-parsimony.md

You are reviewing my .claude.json cleanup tooling.

Context:

• Python script: cleanup_claude_json.py (backs up ~/.claude.json, analyzes projects, and removes entries whose directories no longer exist, with a dry-run/execute flag).
• Strategy doc: CLAUDE_JSON_CLEANUP_STRATEGY.md (describes goals, risks, and a conservative cleanup process).

Tasks (be brief and concrete):

1. In 3–5 sentences, restate the core goal of this script + strategy and the main safety mechanisms (backups, dry-run, scope of deletions).
2. Evaluate the approach against best practices for:
   • config/safety (backups, rollback, blast radius),
   • CLI UX (flags, defaults, messages),
   • maintainability (readability, complexity, future changes).
3. Identify only the few most important improvements (max 5), if any, that would materially increase safety, clarity, or simplicity. Avoid nitpicks.
4. Apply parsimony: point out any part of the design or code that is overcomplicated for the problem and how to simplify it.
5. Decide explicitly: Is this “good enough” for a personal, local tool?
   Answer with: GOOD ENOUGH AS-IS or NEEDS CHANGES, and give a 3–item bullet list explaining why.

Keep the whole response under ~400 words. Do not restate large chunks of code or doc; focus on judgment, tradeoffs, and concrete, high-leverage suggestions.