rules

Roam architecture rule packs

Starter .roam-rules.yml packs tuned for the most common AI-generated-code anti-patterns. Each pack is opinionated but conservative — every rule has a description so the PR comment surfaces why it fired.

Packs

Path	Language(s)	Rules	Severities
python/.roam-rules.yml	Python	14	7 BLOCK, 7 WARN
typescript/.roam-rules.yml	TS / JS / JSX / Vue	14	5 BLOCK, 9 WARN
go/.roam-rules.yml	Go	12	3 BLOCK, 9 WARN
java/.roam-rules.yml	Java	12	5 BLOCK, 7 WARN
kotlin/.roam-rules.yml	Kotlin (Android + Spring)	12	4 BLOCK, 8 WARN
rust/.roam-rules.yml	Rust	12	3 BLOCK, 9 WARN

Usage

Drop a pack at .roam/rules.yml in your repo root:

cp templates/rules/python/.roam-rules.yml .roam/rules.yml
roam rules-validate .roam/rules.yml --explain

Then roam pr-analyze auto-loads it on every PR. Pair with --gate in CI to fail builds on BLOCK-severity violations.

Customising

The packs are intentionally generic — fork them and tune to your codebase:

• Tighten source_glob to your real layout (e.g. apps/api/**/*.py not src/**/*.py)
• Demote BLOCK → WARN for rules you want as advisory
• Add custom forbidden_target_glob patterns for your internal banned APIs

After editing, validate before committing:

roam rules-validate .roam/rules.yml --strict --gate

Pattern types (the four matchers)

• import_from — Python from X import / import X, JS/TS import ... from "X"
• function_call — any call name( or ns.name( (e.g. eval, pickle.loads)
• class_inherit — base classes in class Foo(Base, ...) declarations
• decorator_use — decorator lines @name or @ns.name

Run roam rules-validate --explain for examples + glob syntax.