The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and rewrite arbitrary Polish text drafts, which presents an attack surface for instructions embedded within the data.
Ingestion points: The agent reads external Markdown files and text drafts (draft .md) to perform linguistic analysis and editing.
Boundary markers: The instructions lack specific requirements for using delimiters (e.g., XML tags or triple backticks) to isolate untrusted input from the agent's core instructions.
Capability inventory: The skill utilizes Read, Write, and Edit tools to modify files on the local system.
Sanitization: No specific sanitization or filtering mechanisms are described for the input text before it is processed by the agent.
[SAFE]: All external references and attributions point to legitimate research and well-known software repositories.
The skill references official GitHub repositories for blader/humanizer and deepseek-ai/deepseek-harness for logic and pattern attribution.
The skill mentions the vendor's own domain (matematic.co) for knowledge base updates, representing standard vendor functionality.
No unauthorized network operations, credential access, or persistence mechanisms were detected.

humanizer-pl