refactor-chain
Pass
Audited by Gen Agent Trust Hub on Jul 6, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local JavaScript files (
diagnose.mjs,orchestrate.mjs,ship-gate.mjs) using Node.js to manage the refactoring pipeline's state and logic. These scripts are located within the platform's local skill directory. - [COMMAND_EXECUTION]: The skill invokes project-specific test commands (
--cmd "<test cmd>") and standard version control operations (git stash, commit, PR creation) as part of its 'self-healing' and 'ship' phases. - [DATA_EXPOSURE]: The skill creates and maintains a local state directory (
.refactor-chain) within the user's project to store execution logs, audit trails, and persistent run data, which is standard for long-running developer tools. - [SAFE]: The instructions incorporate multiple safety-first architectural patterns, such as the 'adversarial-verify' step which requires the agent to attempt to break its own 'it worked' claims before advancing. It also includes an 'ADHD contract' that ensures the agent only presents one decision at a time and avoids overwhelming the user with raw technical failures.
- [SAFE]: The skill uses a deterministic 9-phase pipeline that refuses to skip verification steps or proceed past 'gates' (like the security and performance review gate) unless specific pass criteria are met.
Audit Metadata