/upfront:enlighten
What it does
Section titled “What it does”/upfront:enlighten audits your project’s AI instruction files (CLAUDE.md, AGENTS.md) and improves them with stack-specific examples and the three-tier boundary pattern (always / ask first / never).
When to use it
Section titled “When to use it”- First time setting up AI in a repo
- After choosing a tech stack — enlighten generates conventions and examples specific to your stack
- When AI keeps making the same mistakes — add a rule to prevent it
- Periodically as the project evolves
This skill is suggested automatically when /upfront:explore or /upfront:feature detects missing or weak instruction files.
How it works
Section titled “How it works”- Find instruction files — checks for CLAUDE.md, AGENTS.md, .cursorrules, etc.
- Detect tech stack — from package.json, go.mod, pyproject.toml, etc.
- Audit six areas — commands, project structure, code style, testing, git workflow, boundaries
- Present gaps — scored as Present / Missing / Weak
- Generate improvements — stack-specific examples showing good and bad patterns
- Write to file — adds to existing file, never overwrites
The six areas
Section titled “The six areas”Research from 2,500+ repositories shows these are what separate instruction files that work from those that don’t:
- Commands — exact build/test/lint commands with flags, near the top
- Project structure — where code goes, one sentence per directory
- Code style — with good AND bad examples in your language
- Testing — framework, patterns, what to mock and what not to
- Git workflow — commit format, branch naming
- Boundaries — always do / ask first / never do
Key principle
Section titled “Key principle”One real code snippet showing your style beats three paragraphs describing it. Every suggestion includes a concrete, stack-specific example.