dead-code

Warn

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill includes instructions to access platform-specific configuration directories (~/.claude/projects/) and environment files (.env). Writing telemetry data to paths in the user's home directory outside the active workspace is a privacy and security concern.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its broad analysis of untrusted project files combined with destructive capabilities.
  • Ingestion points: Reads all source code, configuration manifests (package.json, requirements.txt), and environment files (.env).
  • Boundary markers: None. There are no instructions to use delimiters or ignore instructions embedded in the analyzed data.
  • Capability inventory: Performs file deletion (Delete entirely dead files), directory cleanup, and arbitrary shell command execution.
  • Sanitization: None. The agent deletes code based on its autonomous interpretation of file contents without human review or validation.
  • [COMMAND_EXECUTION]: The skill executes arbitrary project-defined build and test commands (e.g., npm run build, flutter analyze). This creates a risk where malicious scripts defined in a project's manifest could be triggered during the skill's verification phase.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 23, 2026, 10:58 AM
Security Audit — agent-trust-hub — dead-code