speckit-verify-tasks

To verify completed tasks in a spec-kit project, validate the prerequisites, parse the completed tasks, and check file existence, git-diff relevance, symbol matching, and dead-code wiring to ensure tasks are truly implemented.

A phantom task completion occurs when a task is marked completed but lacks actual implementation. Detect it by checking file existence, validating code symbols are referenced and not dead, and confirming semantic behavior matches expected outcomes.

Dead code detection for completed tasks checks that declared symbols are imported, referenced, and used somewhere in the codebase rather than existing only as stubs, ensuring the implementation is live and wired correctly.

Yes, task verification supports optional task ID filtering and scope selection using branch diffs, uncommitted changes, or plan-anchored commit ranges to limit the verification to the relevant timeframe and changes.

Generate an immutable audit report by running verification with your target scope, producing a machine-parseable report with per-task verdicts including VERIFIED, PARTIAL, WEAK, NOT_FOUND, or SKIPPED results for each completed task.

A PARTIAL or WEAK verdict indicates the completed task has incomplete evidence, meaning some referenced files, code symbols, or expected behaviors are missing, insufficiently matched, or not fully connected within the codebase.