suede-ops-architecture
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of Markdown instructions and YAML/JSON metadata. It does not include any executable scripts (Bash, Python, JavaScript) or binary files.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external system descriptions provided by the user. It contains robust mitigation patterns, including 'Halt' formats that force the agent to stop and request user input for ambiguous data, and a 'Boundaries' section that explicitly prohibits the agent from inventing data or executing build actions.
- [DATA_EXFILTRATION]: No network operations or external URLs for data transmission were detected. The skill operates locally within the agent's context using user-provided information.
- [METADATA_POISONING]: The metadata fields in SKILL.md and CARD.md are descriptive and consistent with the skill's stated purpose of providing architectural guidance.
Audit Metadata