The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process data from the internet, which introduces a surface for indirect prompt injection. Malicious instructions embedded in the HTML or JSON of a fetched website could attempt to influence the agent's subsequent actions.
Ingestion points: Content retrieved via curl and WebFetch as described in SKILL.md.
Boundary markers: Absent; the skill does not instruct the agent to wrap external content in delimiters or provide warnings to ignore embedded instructions.
Capability inventory: The skill provides access to network operations (curl), local command execution (python3, jq), and limited file system interaction (writing response files and cookies).
Sanitization: None; the skill focuses on raw data retrieval and parsing.
[COMMAND_EXECUTION]: The skill relies on shell commands including curl, python3, and jq. While these are used for their intended purpose of web fetching and data processing, they represent significant execution capabilities on the host system.
[EXTERNAL_DOWNLOADS]: The core functionality involves making network requests to external domains. This is expected behavior for a web-fetch skill but involves interacting with untrusted third-party infrastructure.

web-fetch