The Agent Skills Directory

[SAFE]: No code or scripts are included with this skill. It consists entirely of natural language instructions for legal analysis.
[SAFE]: The skill instructs the agent to verify legal information using well-known, official German legal databases (such as gesetze-im-internet.de and dejure.org), which is standard practice for legal research.
[PROMPT_INJECTION]: No patterns associated with prompt injection, role-play bypass, or instruction overrides were detected. The instructions are professional and task-oriented.
[DATA_EXFILTRATION]: No hardcoded credentials, sensitive file paths, or unauthorized network communication patterns were identified.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data (contracts, documents, and screenshots). While it lacks explicit boundary markers or sanitization instructions, it also possesses no dangerous capabilities (like file system writes or network exfiltration) that would make it vulnerable to exploitation.
Ingestion points: External documents, contracts, and screenshots (mentioned in SKILL.md under 'Fallweichen').
Boundary markers: Absent.
Capability inventory: No tool usage or external commands are defined.
Sanitization: Absent.

datenerhebung-zahlen-schwellen-und-berechnung