The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves analyzing external documents such as contracts, evidence files, and logs provided by the user. This represents an attack surface where a malicious document could theoretically contain instructions designed to influence the agent's output.
Ingestion points: Contract documents, evidence files, and user-provided logs mentioned in the 'Sofortstart' and 'Arbeitsweg' sections.
Boundary markers: The skill does not define explicit technical delimiters for external text but instructs the agent to verify claims against official sources and mark uncertain values.
Capability inventory: The skill is limited to text generation, risk assessment, and legal research via authorized tools. It does not perform dangerous file system writes or unauthorized network operations.
Sanitization: The 'Qualitätsfilter' section mandates that uncertain values be marked and linked to missing facts, which serves as a procedural guardrail against manipulation.
[SAFE]: The skill uses official and well-known legal databases for live verification of norms and case law, including gesetze-im-internet.de, dejure.org, and eur-lex.europa.eu. These are trusted authoritative resources.
[SAFE]: No evidence of credential theft, obfuscation, malicious command execution, or persistence mechanisms was found. The instructions are transparent and aligned with the stated legal analysis purpose.

ki-vertragsklauseln-urhebervertragsrecht