The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided content, such as drafts and emails, for editing and rewriting tasks. It lacks explicit structural delimiters or boundary markers for this input, which technically creates a surface where instructions embedded in the user text could attempt to influence the agent's behavior. However, this is inherent to the skill's primary purpose, and the instructions include robust safeguards directing the agent to strictly maintain source facts and prohibiting it from autonomously publishing or sending content.
Ingestion points: User-provided text is ingested via the main skill instructions and the 'learn-from-what-shipped' reference file.
Boundary markers: No specific delimiters are defined to isolate user content from the agent's internal instructions.
Capability inventory: The skill's capabilities are limited to text editing and rewriting; it is explicitly barred from performing network actions like sending or posting content.
Sanitization: The skill does not implement specific sanitization or filtering for the input text.

copy-humanizer