-
Notifications
You must be signed in to change notification settings - Fork 0
FAQ
DUDA includes a complete Manual Mode in SKILL.md with grep-based analysis commands. Python scripts automate the process but aren't required.
Yes. DUDA analyzes import paths and file structure, not framework-specific APIs. It works with Next.js, React, Vue, Angular, Express, NestJS, and any TypeScript/JavaScript project.
Yes! Run duda init at project start to establish the isolation map early. This prevents contamination from the beginning.
ESLint checks individual import statements against static rules. DUDA maps the entire project topology, understands isolation layers, and blocks operations that would create cross-layer contamination — even when individual imports look fine in isolation.
- SCAN: You know the exact file and want to check if it's safe to import
- SCOPE: You know the feature name but not which files are involved
No. SCOPE discovers files; SCAN analyzes them in detail. They're complementary. You can even pipe SCOPE results into SCAN:
duda scope "auth" --files-only | xargs -I {} duda scan {}When DUDA detects migration keywords in your prompt:
- "use X in Y", "copy from", "bring over", "migrate", "port", "apply X to Y"
It always confirms before proceeding.
Not recommended. The threshold exists to prevent isolation breaches. If you're consistently scoring below 95, it likely means there are genuine issues to resolve.
Boundary Trust has zero tolerance. If the isolation policy doesn't exist, or the destination is already contaminated, or the operation conflicts with the deny list — DUDA will not proceed regardless of other scores.
Every DUDA operation is recorded. After 3+ identical patterns, DUDA skips full analysis and applies the cached strategy directly. After 5+ patterns, processing is near-instant.
rm -rf .duda/memory/This forces DUDA to start fresh on the next run.
Yes. The .duda/ directory is per-project.