The Agent Skills Directory

[SAFE]: The skill provides a structured methodology for evaluating legal content. It does not include executable code, scripts, or instructions that could lead to unauthorized system access or data exfiltration.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and score externally provided legal outputs.
Ingestion points: The skill processes legal deliverables (opinions, memos, or AI responses) in the SKILL.md evaluation instructions.
Boundary markers: There are no explicit instructions for the agent to use specific delimiters when reading the input content, although the output format is strictly defined.
Capability inventory: The skill is limited to the Read tool as per the allowed-tools configuration in the frontmatter, preventing it from performing active tasks or network operations.
Sanitization: The skill does not define specific sanitization routines for the processed legal text. However, as its purpose is purely evaluative, the risk is categorized as low.

ocena-outputu-pl