ip-profile

Pass

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted data from user-provided articles to perform writing style distillation.
  • Ingestion points: User-provided 'historical articles' (samples) are collected in Section II and processed in Section III of SKILL.md.
  • Boundary markers: The distillation prompt utilizes <<<SAMPLE n>>> and <<<END SAMPLE n>>> delimiters to isolate untrusted content.
  • Capability inventory: The skill uses the agent's LLM to generate a voiceSystemPrompt based on the samples, which is then used to influence future writing tasks. No direct shell command execution on the samples is present.
  • Sanitization: The system prompt in SKILL.md includes explicit instructions for the model to treat sample content strictly as data and ignore any embedded directives like 'ignore previous instructions'.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 19, 2026, 01:21 PM
Security Audit — agent-trust-hub — ip-profile