The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process various types of data which represents a theoretical attack surface.
Ingestion points: The skill reads local files (logs, source code, JSON) and remote API responses from 'api.squaads.com' and 'api.github.com' using tools like rg, cat, xh, and http.
Boundary markers: The instructions do not define explicit boundary markers or warnings to the agent to disregard instructions potentially embedded within the data being processed.
Capability inventory: The skill has the capability to modify files (sed -i, save), execute shell commands (fd -x), and perform network operations (xh, http).
Sanitization: There are no explicit data sanitization or validation steps mentioned for the content retrieved from external or local files before it is processed by the tools.
[COMMAND_EXECUTION]: The skill makes legitimate use of command-line utilities for its primary functions.
Includes patterns for using fd -x to execute commands on search results and sed for in-place text replacement.
These behaviors are standard for utility-focused skills and are well-documented within the skill content.

utility-pro