The Agent Skills Directory

[SAFE]: The skill instructions define a purely analytical and drafting workflow for legal compliance. No malicious code, obfuscation, or unauthorized data access patterns were detected.
[EXTERNAL_DOWNLOADS]: The skill references official legal documentation from trusted sources, including the European Union's Law portal (eur-lex.europa.eu), the German Federal Ministry of Justice (gesetze-im-internet.de), and the Federal Office for Economic Affairs and Export Control (bafa.de). These references are used for retrieving regulatory statutes and application guidelines.
[INDIRECT_PROMPT_INJECTION]: The skill acts as an ingestion point for external trade data, including technical specifications and end-user statements, which could theoretically contain instructions designed to influence the agent's compliance determination.
Ingestion points: Technical specifications, end-use statements, and end-user data are ingested via user prompts in SKILL.md.
Boundary markers: The skill does not currently define explicit delimiters or instructions for the agent to ignore potentially malicious embedded content within the technical specifications.
Capability inventory: The skill utilizes agent roles (researcher, drafter, reviewer) to perform legal reasoning, document drafting, and review, but does not specify high-risk automated tools like network exfiltration or file system writing.
Sanitization: No explicit input validation or sanitization logic is provided for the external trade data.
[COMMAND_EXECUTION]: The test.md file contains an example command (python ../../../scripts/eval.py) intended for local evaluation of the skill within a development environment. This is a standard practice for skill testing and does not represent an injection risk.

ausfuhr-dual-use-pruefung