scanning-security

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill implements a strong security posture by strictly limiting the agent's available tools. The allowed-tools configuration ensures that only specific, pre-defined security scanning commands and read-only operations can be executed.\n- [PROMPT_INJECTION]: The skill inherently maintains an indirect prompt injection surface due to its core function of analyzing external repository data.\n
  • Ingestion points: The skill reads configuration files from the .claude/ and .github/ directories during the scanning process.\n
  • Boundary markers: There are no explicit markers defined to separate the untrusted scanner output from the instructions provided to the secondary reviewer agent.\n
  • Capability inventory: The risk of exploitation is significantly mitigated because the orchestrating skill is restricted to a specific set of tools, preventing arbitrary command execution.\n
  • Sanitization: Content extracted from the repository is passed to the reviewer agent without explicit sanitization or filtering steps.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 05:18 AM
Security Audit — agent-trust-hub — scanning-security