comfy-debug-preview
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes nodes that ingest and process potentially untrusted data from external sources, such as image metadata, which creates a surface for indirect prompt injection.
- Ingestion points:
CImageLoadWithMetadataandCMetadataExtractor(SKILL.md) are used to pull metadata from external image files. - Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore instructions embedded within the metadata or processed strings.
- Capability inventory: The skill identifies nodes with file-writing capabilities (
DreamLogFile) and console logging capabilities (CConsoleAny,DreamStringToLog) across its instructions. - Sanitization: There is no mention of sanitizing or validating the content extracted from image metadata before it is logged or processed by other nodes.
- [DATA_EXFILTRATION]: The skill facilitates access to system-level telemetry and performance data.
- Evidence:
CUtilsStatSystemandVRAM_Debug(SKILL.md) are used to monitor and log CPU usage, system RAM, and GPU VRAM telemetry to the console or log files.
Audit Metadata