The Agent Skills Directory

[COMMAND_EXECUTION]: The instructions include several shell command examples using the belt CLI to perform tasks like searching for topics (exa/search), generating images (html-to-image), and posting to social media (x/post-create).
[EXTERNAL_DOWNLOADS]: The skill provides installation commands (npx skills add) and links to remote configuration files hosted on the inference-sh GitHub repository to set up the necessary CLI environment.
[REMOTE_CODE_EXECUTION]: The skill contains a Python script snippet that is designed to be executed in a remote environment using the infsh/python-executor tool. The script uses matplotlib to generate a benchmark chart image.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent context via the output of the belt app run exa/search command.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore or sanitize malicious content embedded within the search results.
Capability inventory: The skill has access to the Bash tool and can trigger remote execution and network operations through the belt CLI.
Sanitization: No validation or filtering is performed on the data returned from the external search API before it is used to generate the blog post.

technical-blog-writing