The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted user data regarding procurement details.
Ingestion points: Data is entered via fields like 'Auftragsgegenstand' (object of contract) and 'Eingaben' in SKILL.md, as well as the 'fact_pattern' in test.md.
Boundary markers: The instructions do not define strict delimiters or include 'ignore embedded instructions' warnings for the data it processes.
Capability inventory: The skill is limited to text-based legal analysis and report generation. It does not contain scripts for file system modification, network communication, or system command execution.
Sanitization: No explicit sanitization, filtering, or escaping of the user-provided text is performed before processing.
[EXTERNAL_DOWNLOADS]: The skill provides links to official legal and judicial databases for regulatory reference.
Evidence: The skill references curia.europa.eu, gesetze-im-internet.de, and eur-lex.europa.eu.
Analysis: These domains are official portals for the Court of Justice of the European Union, the German Federal Ministry of Justice, and EU law, respectively.

vergabe-eu-schwellenwert-pruefung