oma-deepsec

Pass

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads the deepsec CLI and related dependencies using standard package managers (bun, pnpm, npm) from Vercel's official repositories and public registries.
  • [COMMAND_EXECUTION]: The skill executes a variety of deepsec CLI commands to perform its core functions, including initializing workspaces, running vulnerability scans, and exporting reports.
  • [DATA_EXFILTRATION]: As an agent-powered scanner, the tool transmits repository code snippets to trusted AI providers, such as Vercel AI Gateway, Anthropic, and OpenAI, for security analysis.
  • [INDIRECT_PROMPT_INJECTION]: The skill acknowledges the risk of processing untrusted code in repositories or pull requests. It provides specific architectural mitigations, such as a two-job CI workflow to isolate repository-write permissions and a sandbox mode to restrict network egress and protect API keys.
Audit Metadata
Risk Level
SAFE
Analyzed
May 16, 2026, 06:54 AM
Security Audit — agent-trust-hub — oma-deepsec