The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user inputs (email, role, budget, and model lists) and interpolates them into a shell command template for execution.
Ingestion points: User-provided parameters for the new user profile defined in SKILL.md.
Boundary markers: No delimiters or specific instructions are provided to the agent to treat these inputs as literal strings rather than executable code.
Capability inventory: The skill utilizes the Bash tool to execute curl commands.
Sanitization: There is no evidence of input validation or escaping mechanisms to prevent command injection through specially crafted user inputs.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform network operations via curl. It transmits a sensitive administrative credential (LITELLM_API_KEY) within the command context to a LiteLLM proxy endpoint.

add-user