The Agent Skills Directory

[SAFE]: The skill serves as a legal compliance framework for 'Sanktionslisten-Screening' (Sanctions List Screening) according to EU regulations (e.g., VO 269/2014). All referenced links point to official government and judicial domains, including eur-lex.europa.eu, gesetze-im-internet.de, and europa.eu.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, specifically 'Trefferdaten' (hit data such as names and birthdates) and 'Geschäftskontext' (business context). This represents a vulnerability surface for indirect prompt injection if the processed data contains hidden instructions. However, the skill lacks autonomous high-privilege tools, and the workflow focuses on human-in-the-loop drafting and review, which significantly mitigates this risk.
Ingestion points: SKILL.md defines inputs for hit data (Name, DOB, Nationality) and business context.
Boundary markers: No specific delimiters or 'ignore' instructions are provided for the external data inputs.
Capability inventory: The skill is limited to drafting memos and performing legal analysis; it does not contain direct tool calls for file-system modifications or network exfiltration.
Sanitization: No explicit sanitization or validation of the input strings is performed by the instructions.
[COMMAND_EXECUTION]: test.md contains a reference to a local evaluation script python ../../../scripts/eval.py. This is a standard testing pattern for this environment and does not involve downloading or executing remote code from untrusted sources.

sanktionslisten-screening