The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided academic drafts and possesses the capability to edit files and execute a local Python script, creating an attack surface where instructions embedded in a paper could influence agent behavior.
Ingestion points: Academic drafts or file contents provided by the user are read during the core loop defined in SKILL.md.
Boundary markers: The instructions do not define explicit boundary markers or delimiters to isolate user-provided text from the system prompt.
Capability inventory: The skill is authorized to use Write, Edit, and Bash(python *) as specified in the allowed-tools metadata.
Sanitization: No sanitization or filtering logic is present to identify or remove potentially malicious instructions within the paper text before processing.

humanizer-paper