What problem does it solve? Teams adopting spec-driven development often apply heavyweight specification ceremony to trivial changes or skip governance entirely on complex work, and AI-generated code reviews frequently make unsupported claims about catching bugs. This Skill applies proportionate governance: hard multi-constraint tasks get a staged approved baseline with attributable drift review, while easy tasks use lightweight reason-first reasoning. ## Core Features & Use Cases - Task Classification Gate: Runs scripts/spec_governance_gate.py to classify work as easy, medium, or hard and route it to the appropriate governance level. - Staged Baseline Workflow: For hard tasks, requires Spec Kit artifacts with named invariants, an approved baseline, then generation in a fresh step so drift findings cite specific clauses. - Claim Guardrails: Blocks false recall claims (e.g., "specs catch more bugs"), inline spec prompting, and unattributed drift findings, keeping a human Accountable for reconciliation. - Use Case: Before implementing a paywall gate with catalog loading, run the gate script to confirm it is a hard task, approve the spec baseline, then generate code in a separate step with drift findings anchored to invariant clauses. ## Quick Start Ask the assistant to classify your task with the spec governance gate script and follow the staged baseline workflow it prescribes before generating any code.
