security-best-practices

Pass

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection due to its core functionality of reading and interpreting untrusted code files.
  • Ingestion points: The SKILL.md workflow instructs the agent to 'inspect the repo' and 'identify ALL languages and ALL frameworks' by reading the project's source files.
  • Boundary markers: There are no instructions to the agent to distinguish between the code data and potential instructions embedded in comments or strings within the files it reviews.
  • Capability inventory: The skill can write a security report (security_best_practices_report.md) and is authorized to 'perform fixes' on the codebase, which could allow a malicious file to trigger unauthorized code changes.
  • Sanitization: No input validation or sanitization is required for the file content being analyzed.
  • [PROMPT_INJECTION]: The 'Overrides' section in SKILL.md instructs the agent to 'bypass or override these practices' if the user asks and tells the agent to 'not fight with them.' This essentially allows a direct prompt to override the primary security-focused purpose of the skill.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 1, 2026, 09:43 AM