The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the user to install the belt-sh/cli package using npx skills add. It also utilizes belt app run to execute various remote applications for research and fact-checking.
[EXTERNAL_DOWNLOADS]: Fetches installation guidelines from a remote GitHub repository (github.com/inference-sh/skills).
[COMMAND_EXECUTION]: Executes shell commands via the belt CLI, including belt login for authentication and multiple instances of belt app run to interact with external APIs.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process information from external search engines (Tavily and Exa), which provides an attack surface for indirect injection.
Ingestion points: Data retrieved from tavily/search-assistant, exa/search, and exa/answer apps.
Boundary markers: None explicitly defined in the prompt to separate external data from internal instructions.
Capability inventory: Ability to execute shell commands restricted to the belt binary.
Sanitization: No instructions provided to sanitize or validate the content retrieved from the search tools before incorporating it into the press release.

press-release-writing