The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads software packages from external registries at runtime to facilitate news monitoring functions. Specifically, it uses uvx to download and execute the web-forager tool and uv run to install the ddgs library from PyPI.
[COMMAND_EXECUTION]: The skill utilizes shell commands to perform news searches and fetch web content. This includes executing uvx web-forager news for search functionality and using curl to retrieve article content via the Jina Reader service.
[REMOTE_CODE_EXECUTION]: The skill downloads and executes code from remote sources at runtime. This is observed in the runtime installation and execution of web-forager via uvx, and the installation of the ddgs library using uv run which executes an inline Python script to handle search fallbacks.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from the web.
Ingestion points: Untrusted data enters the agent context through news search snippets and full-text article fetches from external URLs using the fetch tool and Jina Reader.
Boundary markers: The skill does not define explicit delimiters or special instructions to ignore or isolate embedded malicious prompts within the fetched web content.
Capability inventory: Across its instructions and referenced tools, the skill has the ability to execute shell commands, manage packages, and perform network operations.
Sanitization: There is no evidence of sanitization, filtering, or validation of the content retrieved from external sources before it is summarized for the user.

news-monitor