The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to its core function of processing and transforming untrusted text.
Ingestion points: The skill ingests untrusted data from two primary sources: the text provided by the user for humanization and the contents of files provided for 'Voice Calibration' (specified in SKILL.md and SKILL.pt-BR.md).
Boundary markers: The instructions lack explicit boundary markers, such as XML tags or unique delimiters, to separate the agent's instructions from the text being humanized. There are no warnings to the model to ignore instructions that might be embedded within the user's input.
Capability inventory: The skill is configured with powerful file-system tools including Read, Write, Edit, Grep, and Glob. An injection embedded in a writing sample could attempt to trick the agent into using these tools to read sensitive files or modify system configuration.
Sanitization: The skill does not implement any validation, filtering, or escaping of the input text before it is interpolated into the agent's context.

humanizer