skills/trailofbits/skills/genotoxic/Gen Agent Trust Hub

genotoxic

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading and installing numerous third-party tools and frameworks from external sources including GitHub, Cloudsmith, NPM, and PyPI.\n
  • Evidence: Instructions in references/mutation-frameworks.md for tools like Mull, Stryker, cargo-mutants, and various Go mutation frameworks.\n- [REMOTE_CODE_EXECUTION]: The Mull installation documentation includes patterns for downloading and executing shell scripts directly from a remote server using shell piping.\n
  • Evidence: curl -1sLf 'https://dl.cloudsmith.io/public/mull-project/mull-stable/setup.deb.sh' | sudo -E bash in references/mutation-frameworks.md.\n- [PRIVILEGE_ESCALATION]: The skill documentation includes instructions to use sudo for installing tools, moving binaries to system directories, and modifying system limits.\n
  • Evidence: Commands such as sudo apt-get install, sudo mkdir, sudo cp, and sudo rpm -i are used in the C/C++ Mull installation section of references/mutation-frameworks.md.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from target codebases and external tool outputs to perform its core triage logic and generate reports.\n
  • Ingestion points: The skill parses results from mutation frameworks (mutmut, Stryker, cargo-mutants, slither-mutate) and necessist output to classify findings.\n
  • Boundary markers: No explicit delimiters or boundary markers are defined for the parsing of external JSON or XML tool reports.\n
  • Capability inventory: The skill has the ability to execute shell commands via uv run, install packages via cargo install, and write files (GENOTOXIC_REPORT.md) to the local system.\n
  • Sanitization: There is no detailed evidence of sanitization or strict schema validation for the external tool outputs before they are processed for triage and reporting.\n- [COMMAND_EXECUTION]: The skill uses multiple command-line tools to perform its analysis, including trailmark for graph building and various mutation testing engines.\n
  • Evidence: Quick Start and Workflow sections in SKILL.md detail the execution of uv run trailmark, mutmut run, and necessist.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 01:16 PM
Security Audit — agent-trust-hub — genotoxic