accountable-engineering

# Accountable Engineering Accountable engineering means using AI for leverage while keeping architectural, security, product, and operational decisions understandable and human-owned. Follow this workflow for every non-trivial AI-assisted task. Apply each step directly, loading a named companion skill only when its stated branch applies. This prevents cognitive surrender: accepting generated decisions that nobody can independently explain. ## Workflow ### 1. Define behavior and constraints Write the expected behavior, non-goals, security boundaries, performance expectations, and verification criteria before editing code. **Complete when**: Each requested behavior has a checkable outcome, and every known constraint or non-goal is explicit. ### 2. Propose approach before implementation Inspect the relevant code and propose the smallest approach that fits its existing boundaries. Include data flow, integration points, failure handling, meaningful alternatives, and unanswered questions. Load `blindspot-pass` for hidden gotchas, `spec-interview` when requirements can change the design, or `context-discovery` when the behavior spans multiple modules or tools. **Complete when**: The proposal accounts for every affected boundary and identifies every decision that could change the implementation. ### 3. Review architecture and key decisions Present material architectural, security, product, and rollout choices for review. State a recommendation and its trade-offs for each u...