The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a Python script named hub_init.py from the skill's local directory and runs shell commands provided by the user (or generated by the agent) via the --eval parameter to establish performance baselines.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it interpolates untrusted task descriptions and evaluation commands into shell execution templates. • Ingestion points: Untrusted data enters the context via the --task and --eval arguments described in SKILL.md. • Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious content embedded in the task or eval strings. • Capability inventory: The skill has the ability to execute arbitrary shell commands (via the evaluation parameter) and write to the file system (creating the .agenthub/ directory). • Sanitization: The skill instructions do not specify any validation, escaping, or filtering for the external content before it is used in command execution.

hub-init