kata-ex-hunt-dead-code
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides a technical and legitimate tool for Elixir code auditing. The analysis script
scripts/dead_code.exsis a sophisticated AST walker that handles Elixir-specific constructs like macros and protocols without any hidden malicious logic.\n- [DYNAMIC_EXECUTION]: The script usesCode.eval_fileto load configuration from theroots.exsfile. In the Elixir ecosystem, this is the standard method for loading configuration and build manifests (similar tomix.exs). The manifest file is sourced from the local project or a bundled template, representing a controlled use of dynamic execution for configuration purposes.\n- [INDIRECT_PROMPT_INJECTION]: As the tool ingests and analyzes source code from a project, it presents an indirect prompt injection surface. A malicious source file could potentially include text designed to influence the agent's interpretation of the dead-code report.\n - Ingestion points: The script reads Elixir source files (
.ex,.exs) from the target project'slib/,test/, andconfig/directories.\n - Boundary markers: The results are presented in a structured text or JSON format, and the
SKILL.mdinstructions explicitly require the agent to verify findings before taking action.\n - Capability inventory: The agent is authorized to perform file read, write, and delete operations on the project code based on the tool's output.\n
- Sanitization: The script extracts specific metadata (module and function identifiers) through AST parsing, which naturally filters out arbitrary string content from the source files, reducing the risk of injection instructions reaching the final report.
Audit Metadata