oma-deepsec
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches and executes the deepsec vulnerability scanner and Claude Code CLI from trusted registries using bunx, pnpm, and npm. The tools are maintained by Vercel Labs and Anthropic, respectively.- [COMMAND_EXECUTION]: Invokes deepsec CLI primitives to manage scanning lifecycles, including workspace initialization, AI-driven processing, and revalidation of security findings.- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external source code from target repositories, creating a potential injection surface. This risk is addressed through detailed guidance on using 'deepsec sandbox' mode for untrusted codebases and providing explicit boundary instructions in INFO.md files.- [SAFE]: Implements a secure two-job CI/CD pattern that prevents PR-controlled code from accessing repository write permissions, significantly reducing the risk of supply chain attacks during automated reviews.
Audit Metadata