llm-citation-monitor
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is purely informational and template-based. It provides structured guidance for monitoring how AI assistants (like ChatGPT or Claude) cite specific websites.
- [COMMAND_EXECUTION]: The
SKILL.mdfile contains validation commands (python3 $CODEX_HOME/skills/...). These are standard development linting tools and do not represent a security risk. - [CREDENTIALS_UNSAFE]: The skill explicitly forbids the storage of credentials, cookies, or API keys in its 'Non-Negotiables' and 'Privacy' sections, adhering to security best practices.
- [DATA_EXFILTRATION]: No network-enabled scripts or exfiltration patterns were detected. The skill focuses on manual observation and reporting.
- [PROMPT_INJECTION]: The
evals.jsonfile includes negative test cases to ensure the agent refuses to store credentials or make unsupported claims, reinforcing safety boundaries.
Audit Metadata