anfechtung-beschluss
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze user-provided documents to identify legal errors in resolutions. This represents an ingestion point for untrusted data. However, the skill does not grant the agent any dangerous capabilities (such as network access, file writing, or command execution) that could be exploited via embedded instructions in those documents. The risk is limited to the agent potentially following instructions within the user document instead of the skill's guidelines, which is a standard LLM behavioral risk rather than a programmatic vulnerability.
- [DATA_EXPOSURE]: The skill contains a 'Vorsichtsregel' (precautionary rule) that explicitly directs the agent to minimize the collection of sensitive data, such as health, financial, or family information, and to only provide necessary and verifiable facts to authorities or opponents. This is a positive security practice.
- [NO_CODE]: The skill consists entirely of Markdown instructions and does not include any scripts, configuration files, or tool definitions that could execute code on the host system.
Audit Metadata