codebase-pruner

Pass

Audited by Gen Agent Trust Hub on Apr 25, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill's instructions prioritize repository integrity by requiring a clear dead-code inventory and reference mapping before any modifications are made. Hard rules prevent the deletion of dynamically-referenced code or generated files.\n- [COMMAND_EXECUTION]: The skill suggests running local verification commands such as 'npm run lint', 'type-check', and 'targeted tests' to ensure stability after each deletion slice. These commands are standard for codebase maintenance and are not used for malicious purposes.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes untrusted data from the entire codebase during the audit and pruning process.\n
  • Ingestion points: Source code, configuration files (CI/CD, Docker), and package exports throughout the repository.\n
  • Boundary markers: None present to distinguish codebase data from instructions.\n
  • Capability inventory: The skill can read, modify, and delete local files, and execute predefined verification scripts via the shell.\n
  • Sanitization: No sanitization of the scanned content is performed; safety relies on the agent's logic, confidence scoring, and manual review of uncertain candidates.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 25, 2026, 10:57 AM
Security Audit — agent-trust-hub — codebase-pruner