What problem does it solve?
When using spec-driven development, it is common for specs to be marked as complete even when implementation is partial, missing, or misaligned with documented requirements. Manual checks are time-consuming, prone to human error, and often lack concrete evidence to back up claims. This skill eliminates that guesswork by performing a rigorous, evidence-based audit of implemented specs against the actual codebase, giving you an objective, verifiable report of what is actually built.
Core Features & Use Cases
- Constraint Evidence Auditing: Walks every tagged constraint (ARCH, PERF, KD, etc.) in the spec and grounds each verdict in concrete file:line evidence from the codebase, classifying constraints as Implemented, Partial, Unverified, or N/A.
- Acceptance Command Validation: Executes the spec's defined acceptance commands, uses git stash round-trips to distinguish pre-existing failures from failures introduced by the spec implementation, and records full output and duration for each command.
- Standardized Audit Reporting: Generates a consistent, structured audit document in docs/audits/ covering structural metadata checks, constraint coverage, acceptance results, and severity-classified issues, with no modifications to the original spec files.
- Use Case: For engineering teams using the /spec format for spec-driven development, run this skill after implementing a spec to get an objective closure report before marking the spec as done, catching missing features, broken tests, and spec-code drift early.
Quick Start
Use the validate-spec skill to audit the implementation of your target spec ID and generate a complete evidence-based validation report in your project's docs/audits/ directory.