gap-foerderantrag

Pass

Audited by Gen Agent Trust Hub on Jun 28, 2026

Risk Level: SAFE
Full Analysis
  • [DATA_EXPOSURE]: The skill implements a privacy safeguard by explicitly instructing users to redact sensitive identifiers like the InVeKoS business number (using placeholders like [Betr.-Nr.]) and providing field lists as anonymized data.
  • [EXTERNAL_DOWNLOADS]: The skill references official legislative and judicial resources. These include EU law (eur-lex.europa.eu), German federal laws (gesetze-im-internet.de), and official court databases. These are trusted government domains and do not involve executable code downloads.
  • [PROMPT_INJECTION]: The skill processes external data such as legal notices and business data to generate legal drafts (Indirect Prompt Injection surface).
  • Ingestion points: User-provided fact patterns, field lists, and prior administrative notices (SKILL.md).
  • Boundary markers: The skill uses a structured 'Gutachten' (legal opinion) template to separate logic from data.
  • Capability inventory: No subprocess calls or network operations are present; capabilities are limited to text analysis and drafting.
  • Sanitization: Proactive instructions for data redaction are present.
  • [COMMAND_EXECUTION]: The test.md file contains a local script execution command (python ../../../scripts/eval.py) for developer testing. This is a standard project-relative path and does not fetch or execute remote code.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 28, 2026, 08:40 AM
Security Audit — agent-trust-hub — gap-foerderantrag