The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool with a restriction to the belt command (belt *) in the allowed-tools frontmatter. This scopes the agent's interaction to the specific CLI tool required for the service.
[EXTERNAL_DOWNLOADS]: The instructions guide users to install the belt CLI via npx skills add belt-sh/cli and point to installation instructions on GitHub (inference-sh/skills). These are standard operations for enabling the skill's functionality within the inference.sh ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill's research and extraction workflows ingest data from untrusted external web pages and provide that data to an LLM.
Ingestion points: External content from web searches and extraction is stored in local files (search_results.json, content.json) in SKILL.md.
Boundary markers: The skill suggests using XML-style tags such as <search-results> and <content> to separate external data from system instructions.
Capability inventory: The skill utilizes the Bash tool (restricted to belt) to fetch the data.
Sanitization: There is no evidence of filtering or sanitization being applied to the external content before it is processed by the agent.

web-search