The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches content from arbitrary URLs using requests and BeautifulSoup to compile research reports. This creates a surface where malicious instructions embedded in web pages could influence the agent's behavior during synthesis or report generation.
Ingestion points: Content is fetched in the extract_page_content function within SKILL.md.
Boundary markers: None identified. The instructions do not tell the agent to ignore instructions embedded in the fetched text.
Capability inventory: The skill uses requests for network access, pip for package installation, and shell commands (cat > file) for file writing.
Sanitization: The skill removes some HTML tags (nav, footer, etc.) but performs no sanitization or escaping of the actual text content before it is processed by the LLM.
[COMMAND_EXECUTION]: The skill uses shell commands to install dependencies (pip install requests beautifulsoup4) and save reports (cat > research_report_[topic].md << 'EOF'). These are standard operational commands for the skill's purpose.
[EXTERNAL_DOWNLOADS]: The skill fetches data from the Brave Search API (api.search.brave.com) and arbitrary external URLs during the research process. These network operations are intrinsic to the primary research functionality.

web-research