What problem does it solve? After a feature is implemented in a spec-driven pipeline, teams need a verifiable record mapping every Success Criterion and Functional Requirement in spec.md to a verification approach — without the orchestrator re-reading large artifacts and polluting its context. This Skill closes that gap by dispatching an isolated tester session and enforcing a conformance gate before the phase can be called complete. ## Core Features & Use Cases - Isolated tester dispatch: Sends exactly one Sonnet subagent whose only inputs are completion-report.md and spec.md (with an optional lazy read of implement.log.md), keeping the main session's context clean. - Conformance gating: Runs check-conformance.py against the feature directory and blocks the phase on any finding scoped to testing.md, relaying findings verbatim and restoring any prior valid file. - Auditable tracing: Appends exactly one trace record to traces.jsonl with role tester, model claude-sonnet-5, token capture, and a context_in field listing the files the tester actually read. - Use Case: After running /speckit-complete on a feature, invoke /speckit-testing to produce testing.md mapping each requirement to a verification approach, mark uncovered items as explicit GAPs, and optionally trigger the after_testing git commit hook. ## Quick Start Run /speckit-testing in a repository with Spec Kit initialized after /speckit-complete has produced a valid completion-report.md for the current feature.
