agently-agentic-design
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill package contains only instructional Markdown documentation and a YAML configuration file. There are no executable scripts (.sh, .py, .js), binaries, or automation files that perform actions on the host system.
- [SAFE]: All terminology and framework references (such as TriggerFlow, TaskWorkspace, and RecordStore) are consistent with the vendor's (agentera) architectural ecosystem and are used here for design guidance rather than implementation.
- [SAFE]: The instructions explicitly advocate for security-first design principles, including mandatory redaction of secrets, fail-closed logic for unauthorized access, and rigorous validation of data provenance.
- [SAFE]: Network-related mentions (such as referencing Anthropic and Google engineering blogs or ArXiv papers) are used strictly for academic and professional documentation purposes and do not trigger unauthorized connections.
- [SAFE]: The skill demonstrates high awareness of indirect prompt injection and exfiltration risks, providing specific 'Non-Negotiable Boundaries' and 'Completion Gates' to help agents design safer systems that avoid these common pitfalls.
Audit Metadata