What problem does it solve? After a spec-driven implementation run finishes, there is no trustworthy record of what actually happened unless someone writes it down accurately. This Skill closes out an implement run by re-reading tasks.md and implement.log.md from disk on every invocation and authoring a validated completion-report.md, so the pipeline's historical record reflects what is on disk rather than what the model happens to remember. ## Core Features & Use Cases - Disk-grounded reporting: Re-reads tasks.md and implement.log.md in full on every run (rule R1-S02), never authoring the report from retained context, so reports stay accurate across compactions, clears, and resumed sessions. - Contract-validated output: Writes completion-report.md in the exact format docs/contracts/completion-report.md fixes — frontmatter status in {success, partial, failed} plus seven ordered core sections ending in a Telemetry section. - Telemetry rollup: Computes per-phase and total token/duration figures via cost_evidence.py and phase_scope.py, relaying values verbatim with FULL / PARTIAL (lower bound) / NO SIGNAL / UNAVAILABLE labels. - Use Case: After running /speckit-implement-parallel on a feature, invoke /speckit-complete to produce the completion report that downstream phases (like the testing doc's coverage map) and hooks (after_complete) depend on. ## Quick Start Run /speckit-complete with no arguments after an implement run to generate the current feature's completion-report.md from its tasks.md and implement.log.md.
