The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides instructions for the agent to adopt the persona and cognitive framework of Naval Ravikant. These instructions are focused on character simulation and do not contain patterns intended to bypass safety guidelines, extract system prompts, or override administrative constraints.\n- [DATA_EXFILTRATION]: No attempts to access sensitive files (such as credentials or environment variables) or exfiltrate data to external domains were identified. The skill's operation is confined to its stated purpose.\n- [REMOTE_CODE_EXECUTION]: No unauthorized or suspicious remote code execution patterns, such as piping network downloads to a shell, were found.\n- [SAFE]: The skill references standard web search tools to classification and research topics (as outlined in Step 2 of the Answer Workflow in SKILL.md). While this constitutes an ingestion of external data (Indirect Prompt Injection surface), it is a standard functional requirement for research and does not involve exploitable capabilities. \n
Ingestion points: WebSearch results in SKILL.md.\n
Boundary markers: The skill instructs the agent to create an internal summary of facts before responding.\n
Capability inventory: WebSearch utility.\n
Sanitization: Not explicitly defined, but the workflow focuses on fact verification.

naval-perspective