The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bash script within SKILL.md that utilizes standard utilities including jq, mv, and echo to modify local configuration files.
[PROMPT_INJECTION]: The skill functions as an ingestion point for user-defined strings that are later rendered as guidance to influence the agent's behavior, establishing an indirect prompt injection surface.
Ingestion points: User input is captured via the ARGUMENTS environment variable in SKILL.md.
Boundary markers: No delimiters or explicit warnings to ignore embedded instructions are used for the stored forbidden strings.
Capability inventory: The script in SKILL.md performs file system writes and state updates using jq and mv.
Sanitization: The implementation correctly utilizes jq --arg to ensure user-supplied data is treated as a literal string within the JSON structure, effectively preventing direct command or schema injection.

forbid