deduplicate

Pass

Audited by Gen Agent Trust Hub on Aug 3, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted codebase data and has the capability to execute commands via project test suites and linters. This presents a potential surface for indirect prompt injection where malicious instructions in a repository could influence the agent's behavior during analysis.
  • Ingestion points: first-party codebase files (SKILL.md)
  • Boundary markers: None specified to delimit untrusted code from instructions
  • Capability inventory: executes project-specific formatters, linters, typecheckers, and test suites (SKILL.md)
  • Sanitization: No explicit sanitization or filtering of codebase content is mentioned
  • [COMMAND_EXECUTION]: The skill includes a verification step that invokes existing local development tools. This is standard functionality for refactoring tools to ensure code integrity.
  • [DATA_EXPOSURE]: The skill performs broad searches across the codebase to identify duplication. This activity is restricted to first-party code and is necessary for the skill's primary function.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 3, 2026, 09:06 PM
Security Audit — agent-trust-hub — deduplicate