The Agent Skills Directory

DATA_EXFILTRATION (LOW): The script scripts/research_lookup.py performs outbound network requests to openrouter.ai. While this domain is essential for the tool's core research functionality, it is not on the trusted whitelist for network operations.
PROMPT_INJECTION (LOW): The skill implements an indirect prompt injection surface by interpolating user-controlled data into LLM prompts and displaying the output. 1. Ingestion points: User queries captured from CLI arguments in lookup.py and the query parameter in ResearchLookup.lookup. 2. Boundary markers: The query is wrapped in simple double quotes within the _format_research_prompt template, which is insufficient to prevent instruction override. 3. Capability inventory: Network access via requests.post to external LLM providers through OpenRouter. 4. Sanitization: There is no sanitization of the user-provided query nor any validation/filtering of the content retrieved from the external API before presentation.

research-lookup