The Agent Skills Directory

[PROMPT_INJECTION]: No override instructions, safety bypasses, or system prompt extraction attempts were found. The instructions are strictly limited to the professional domain of AI governance.
[DATA_EXFILTRATION]: No access to sensitive files or hardcoded credentials was detected. All external URLs point to legitimate, official legal and regulatory domains (e.g., eur-lex.europa.eu, gesetze-im-internet.de).
[OBFUSCATION]: No use of Base64, zero-width characters, homoglyphs, or other encoding techniques to hide intent was found in the skill or test files.
[REMOTE_CODE_EXECUTION]: No patterns involving remote script downloads (curl/wget | bash) or dynamic code execution (eval/exec) were identified.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied information regarding AI system contexts. While this constitutes an attack surface (ingestion points: system metadata, context descriptions), the skill's capabilities are limited to document drafting, posing minimal risk. No explicit boundary markers were observed, but the structural template for output mitigates confusion.
[COMMAND_EXECUTION]: The skill does not instruct the agent to execute shell commands. A local python script call found in test.md (python ../../../scripts/eval.py) is part of the repository's internal development and testing workflow and is not an indicator of malicious intent.

ki-monitoring-konzept