The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses the ${ARGUMENTS} placeholder to ingest user input, which is then used to generate a new command file. This pattern creates a vulnerability to indirect prompt injection, as malicious input could be designed to override the instructions of the resulting command.\n
Ingestion points: The ${ARGUMENTS} variable in SKILL.md.\n
Boundary markers: None identified; the input is placed directly into the goal section of the generated content.\n
Capability inventory: The skill allows Write and Bash operations.\n
Sanitization: No explicit sanitization or validation of the input content is performed.\n- [COMMAND_EXECUTION]: The skill is granted Bash tool access. Although intended for development tasks like linting and testing, this provides a capability that could be exploited if the agent is misled by malicious instructions.

create-command