Goal-driven autonomous execution with built-in irreversibility guard. Port of Anthropic /goal (Claude Code v2.1.139) that automatically pauses for explicit user approval when detecting unrecoverable operations — git push, repo visibility changes, payment, secret exposure, mass data deletion.
1 skill
11-section rigorous design plan review covering problem framing, scope, architecture, security, observability, deployment, performance, reliability, testing, maintainability, and migration. Adapted from gstack /plan-ceo-review for sole-developer + AI-pair-programming workflows.
1 skill