The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the 'wc -l' system command to determine the line count of files specified in the input. This is a standard part of its documented functionality to ensure tasks do not exceed size ceilings.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted specification files, creating an attack surface for indirect prompt injection. * Ingestion points: Reads the specification file provided as an argument in SKILL.md. * Boundary markers: No explicit delimiters or validation rules are defined for the input content. * Capability inventory: Performs file measurement and writes a single new file to the local directory. * Sanitization: No sanitization is performed on the input file paths or content.

nuke-split