The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses uvx and uv to dynamically download and run external Python packages.
Evidence: uvx --python '>=3.10,<3.14' web-forager search and uv run --with 'ddgs>=9.5.2' python are used to perform web searches when built-in tools are unavailable.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to interact with search engines and fetch web page content.
Evidence: Use of curl -s "https://r.jina.ai/..." to fetch markdown versions of web pages and uvx/uv for search operations.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection because it ingests and processes untrusted data from the web.
Ingestion points: Web search results (web-forager, ddgs) and full page content fetched via Jina Reader or built-in tools (SKILL.md).
Boundary markers: The instructions do not define specific delimiters or warnings for the agent to ignore instructions embedded within the fetched external content.
Capability inventory: The skill has access to shell execution (curl, uv, uvx) and search tools, which are used to process the untrusted data.
Sanitization: No specific filtering or sanitization of the fetched web content is requested before the agent evaluates it.

tech-advisor