The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates untrusted user data into generated documentation templates.
Ingestion points: User input is directly interpolated into the {exact user request} placeholder in the 01-objectives.md template.
Boundary markers: While the template wraps the placeholder in double quotes, it lacks explicit instructions to the agent to disregard any instructions contained within the user-provided text.
Capability inventory: The skill requires the agent to perform file system write operations to scaffold the task folder structure defined in SKILL.md.
Sanitization: There are no sanitization or validation steps identified to filter potentially malicious instructions from the user input before interpolation.
[NO_CODE]: The skill consists solely of markdown templates and configuration, with no executable logic or scripts included.

exploration-protocol