gemini-forge
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data that could potentially contain malicious instructions intended to manipulate the model's output or the agent's actions.
- Ingestion points:
scripts/load_design_system.pyrecursively reads files from a user-specified directory;scripts/screenshot_to_code.pyprocesses image files for technical analysis;scripts/generate_ui.pyreads design context from external text files. - Boundary markers: External content is wrapped in markdown headers and code blocks within the
load_design_system.pyscript to provide structural separation, though this does not fully prevent sophisticated injection attacks. - Capability inventory: The skill is capable of writing multiple files to the local filesystem through its
GeminiForgeClient.save_codeutility and making outbound network requests to the Gemini API endpoint. - Sanitization: The skill does not perform sanitization or filtering of the ingested file content or screenshot analysis text for potential prompt injection patterns before including them in the generation prompts.
Audit Metadata