arc42-section-12
Pass
Audited by Gen Agent Trust Hub on Jul 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is instructional and follows a structured template approach to documentation. It does not include any hidden commands, obfuscated content, or persistence mechanisms.
- [DATA_EXFILTRATION]: The skill instructions involve reading existing project files (arc42 Sections 3, 4, 5, 8, and 9) to extract candidate terms. This access is localized to the project environment and is necessary for its stated purpose. The skill does not use any network tools or command-line utilities to transmit data.
- [EXTERNAL_DOWNLOADS]: The skill references the official arc42 documentation site (docs.arc42.org) for reference purposes. This is a well-known resource in the software architecture community and does not involve any automated downloads or code execution.
- [PROMPT_INJECTION]: The skill contains logic to process data from other files (indirect ingestion). While this presents a theoretical surface for indirect prompt injection if those files contained malicious instructions, the skill lacks the capabilities (such as file writing or network access) to cause harm beyond influencing the generated text of the glossary.
Audit Metadata