The Agent Skills Directory

[SAFE]: The skill's instructions and logic are consistent with its stated purpose. It provides comprehensive guidelines for text editing without any hidden or suspicious functionality.
[PROMPT_INJECTION]: An indirect prompt injection surface is present because the skill ingests untrusted text and has permissions to modify the filesystem using tools like Write and Edit.
Ingestion points: User-provided text for humanization (SKILL.md).
Boundary markers: Absent; the instructions do not use delimiters or specify that input text should be treated strictly as data.
Capability inventory: The agent is granted Read, Write, Edit, Grep, and Glob tools to perform its tasks (SKILL.md).
Sanitization: Absent; the skill relies on the model to identify patterns within the provided text without pre-processing filters.

humanizer