The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data including contracts, administrative protocols, and legal briefs, creating a potential surface for indirect prompt injection. * Ingestion points: Legal documents, administrative acts, and external norm databases (SKILL.md). * Boundary markers: No specific delimiters or instructions are provided to the agent to distinguish between user data and instructions within the processed documents. * Capability inventory: The skill's capabilities are limited to generating risk matrices and memos; no dangerous commands, file writes, or unexpected network operations are present. * Sanitization: The instructions do not define any validation or sanitization steps for the imported documents.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill references and fetches data from well-known legal portals such as gesetze-im-internet.de, dejure.org, and eur-lex.europa.eu. These network operations target official and trusted sources and are appropriate for the skill's stated legal purpose.

vg-bild-vgwort-verlag