The Agent Skills Directory

[PROMPT_INJECTION]: The skill is highly susceptible to indirect prompt injection attacks because it instructs the agent to ingest and analyze untrusted content from the web.
Ingestion points: The skill utilizes agent-browser, web_search, and curl to retrieve data from arbitrary external URLs found during reconnaissance (identified in SKILL.md and references/recon-vectors.md).
Boundary markers: The instructions do not define clear delimiters or warning markers to isolate untrusted external content from the agent's core instructions.
Capability inventory: The skill possesses significant network capabilities through its tools, allowing it to navigate the web and fetch remote artifacts.
Sanitization: No explicit instructions are provided for the agent to sanitize or filter potentially malicious text (e.g., system-level override commands) embedded in fetched social media bios or website metadata.

osint-investigator