boss
Pass
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The orchestrator utilizes a local CLI suite (
boss runtime) and theBashtool to manage development workflows. It implements a safety hook to intercept and block dangerous operations such asrm -rf,git push --force, andDROP TABLE. - [EXTERNAL_DOWNLOADS]: The skill manages its own updates via
npm update @blade-ai/boss-skill. It also utilizesagent-browserfrom Vercel Labs, a trusted organization, to perform automated UI testing during the quality assurance phase. - [PROMPT_INJECTION]: The skill is an orchestration engine that interpolates user requirements into sub-agent prompts. While this presents a surface for indirect prompt injection, the risk is significantly mitigated by mandatory technical reviews, automated code-quality gates (Gate 0), and an 'attack-oriented' QA protocol defined in the sub-agent instructions.
Audit Metadata