ai-asset-pricing: Skill for Claude Code

Rule Creator & Auditor

Two modes: Create a new rule or Audit an existing one.

Examples

• /rule-create -- create a new rule (interactive)
• /rule-create database patterns for PostgreSQL -- create with topic description
• /rule-create audit .claude/rules/validation.md -- audit a specific rule
• /rule-create audit all -- audit all rules with summary table

Mode Selection

Parse $ARGUMENTS:

• Empty, new, or a topic description → Create mode
• audit, audit <path>, or audit all → Audit mode

---

CREATE MODE

Four phases: Gather → Draft → Write → Verify.

Phase 1: Gather Context

1a. Scan existing rules

List all files in .claude/rules/ to understand what rules already exist. This prevents creating duplicates and helps identify the right scope.

1b. Ask the user

If $ARGUMENTS contains a topic description (anything beyond new), use it as the domain. Otherwise, use AskUserQuestion with all of the following:

1. Domain: What does this rule cover? (e.g., database patterns, security, API conventions, testing, writing style, deployment)
2. Scope: What file types or paths should trigger this rule? Give glob patterns (e.g., **/*.py, src/api/**/*.ts) or say "always load" if it applies everywhere.
3. Key constraints: What are the 3-5 most important things Claude must know or do when this rule applies? Be specific — concrete patterns, not abstract goals.
4. Bad patterns: Are there specific anti-patterns or mistakes Claude should avoid? (optional but valuable)

1c. Check for overlap

Compare the user's answers against existing rules. If there's significant overlap with an existing rule, warn the user and ask whether to:

• Extend the existing rule instead
• Create a new focused rule covering only the non-overlapping parts
• Proceed anyway (user's call)

Phase 2: Draft

2a. Choose a template

Based on the domain, select the best body pattern from the reference:

Domain Type	Template Pattern	Best For
Code conventions	Quick Reference + Sections	Naming, formatting, architecture patterns
Security / correctness	Do / Don't / Why / Refs	Vulnerability prevention, data integrity
Writing / style	Banned/Required + Examples	Prose style, documentation standards
Data / schema	Schema + Query + Gotchas	Database conventions, data pipelines