editorial-portrait-composition-by-1f0mjj8
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to access and inspect media hosted on well-known platforms, including
goodcase.ai,youmind.com, andx.com(Twitter), to serve as visual anchors for prompt generation. These references are documented as evidence for the artistic styles and do not involve executable code or sensitive data retrieval. - [INDIRECT_PROMPT_INJECTION]: The workflow involves ingesting external image data and text from
references/cases.md. While this presents a surface for indirect prompt injection (e.g., instructions hidden within image metadata or OCR), the skill's instructions include specific logic for the agent to 'Preserve,' 'Replace,' and 'Avoid' specific traits, which acts as a structured boundary for processing the external content. - [COMMAND_EXECUTION]: The skill uses instructional language such as 'Turn the method into an executable... workflow,' which is a natural language metaphor for creating a repeatable agent process and does not refer to the execution of shell commands or scripts.
- [DATA_EXFILTRATION]: No patterns of data exfiltration or unauthorized credential access were detected. The skill specifically warns against publishing, purchasing, or making external account changes without explicit user approval in its safety section.
Audit Metadata