The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied information such as product types, benefits, and existing text through prompt templates in skill-report.json. While this creates an injection surface, it is inherent to the skill's primary function of editing and generating text.
Ingestion points: User input interpolated into prompt templates (e.g., [product type], [feature name]).
Boundary markers: Absent; user input is directly inserted into the instructional prompt.
Capability inventory: Limited to Write and Edit tools for text generation and modification.
Sanitization: None detected in the provided instructions.
[COMMAND_EXECUTION]: Manual review confirms that no shell commands are present. Words identified as 'risk factors' in the skill's metadata (e.g., 'Deploy', 'Launch', 'Build') are used purely as copywriting terminology and not as executable commands.
[DATA_EXFILTRATION]: No network activity or sensitive data access detected. The skill operates entirely on text provided in the conversation context.

copywriter