The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill processes external CSS and UI component files which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: **/*.css and app/components/ui/** via file_triggers.
Boundary markers: Absent; no explicit delimiters or warnings for the agent to ignore embedded instructions in CSS.
Capability inventory: The skill executes a local shell script ./scripts/log-skill.sh.
Sanitization: Absent; the skill does not specify any sanitization of the file content before processing.
[Command Execution] (LOW): The skill executes a local bash script (./scripts/log-skill.sh) for instrumentation purposes. While intended for logging, any script execution increases the attack surface if the script contents are not properly audited.

mobile