The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user text and has the ability to modify files via the Write and Edit tools. It lacks explicit boundary markers or instructions to ignore embedded commands within the text to be humanized, which could allow an attacker to trigger unintended actions. \n- Ingestion points: User input provided for humanization (SKILL.md) or files read by the agent via the Read tool. \n- Boundary markers: None present in SKILL.md to distinguish data from instructions. \n- Capability inventory: Access to Read, Write, Edit, Grep, Glob, and AskUserQuestion tools (SKILL.md). \n- Sanitization: No sanitization or filtering of input text is performed.

humanizer