qwen-edit
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied images and text prompts to interact with an external AI model, which creates an inherent attack surface for indirect prompt injection.
- Ingestion points: User-provided images and text strings are ingested via the
--inputand--promptarguments of thetools/image_edit.pyscript as described inSKILL.mdandparameters.md. - Boundary markers: The prompting patterns provided in
prompting.mdandexamples.md(e.g., "[Action], [details], [constraints]") do not utilize delimiters, XML tags, or explicit instructions to the model to ignore potentially malicious commands embedded within the input data. - Capability inventory: The skill facilitates command execution of a local script (
tools/image_edit.py) which performs network operations to communicate with the RunPod serverless API for image transformation. - Sanitization: There is no documentation or evidence in the provided files of input validation, prompt escaping, or sanitization to prevent adversarial instructions from influencing the agent or the underlying AI model.
Audit Metadata