-
Notifications
You must be signed in to change notification settings - Fork 1k
Open
Description
Feature Request: Richer Spec Validation & Tooling
Problem Description
Current validation checks are structural (Schema-aware) but lack semantic depth. Specs can be syntactically correct but ambiguous, underspecified, or inconsistent in style.
Proposed Solution
- Schema-aware Validation: Extend validation to semantic scenarios, check for edge case coverage.
- Ambiguous Terms Detection: Warn on terms like "should", "may", "could" in requirements (enforce "SHALL/MUST").
- Underspecified Checks: Warn if requirements lack scenarios or acceptance criteria.
- Automated Quality Rules: Lint specs for completeness and cohesion.
Benefits
- Catches gaps early in the design phase.
- Ensures consistency across large spec suites.
- Reduces "prompt drift" when specs are passed to AI assistants.
Industry Inspiration
- Traycer: "Automated quality rules" - linting specs for completeness and testability.
Related Issues
- Feature request: Include acceptance verification instructions for each task in AGENTS.md prompts #345: Feature request for including acceptance verification instructions. This aligns with our "Underspecified Checks".
Metadata
Metadata
Assignees
Labels
No labels