The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to create research directories (mkdir) and execute internal Python scripts for web scraping and data processing.
[EXTERNAL_DOWNLOADS]: The documentation references the installation of the requests library via pip3, which is a standard and well-known dependency for Python-based web operations.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it fetches and processes content from external websites, such as image alt-text, headings, and social media mentions.
Ingestion points: Untrusted content is ingested from external URLs via WebFetch, WebSearch, and the requests library within the provided Python scripts.
Boundary markers: No specific delimiters or instructions are used to prevent the agent from obeying commands that might be embedded in scraped website content.
Capability inventory: The skill utilizes bash for running scripts and managing files, providing a functional surface if instructions were successfully injected.
Sanitization: Scripts include basic string cleaning to normalize company names (e.g., removing file extensions and title-casing), but do not implement security-focused sanitization to filter out executable instructions.

customer-discovery