reuse-or-replace

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: Analysis of the skill instructions and evaluation cases indicates a benign purpose focused on software development discipline. No indicators of obfuscation, malicious network activity, credential theft, or unauthorized persistence mechanisms were found.\n- [PROMPT_INJECTION]: The skill relies on analyzing external, untrusted content from a project's codebase, which creates a surface for indirect prompt injection.\n
  • Ingestion points: The skill directs the agent to analyze code idioms, module structures, naming shapes, error-handling logic, ADRs, and pull request descriptions across the codebase (SKILL.md).\n
  • Boundary markers: There are no explicit delimiters or specific 'ignore embedded instructions' warnings provided within the skill for the data it processes.\n
  • Capability inventory: The skill's scope is limited to reading, auditing, and reporting on code structure; it does not contain instructions for executing external code, writing to sensitive files, or making network requests.\n
  • Sanitization: No validation or escaping of the ingested code or documentation is specified in the instruction set.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 07:56 PM
Security Audit — agent-trust-hub — reuse-or-replace