flux-best-practices
Pass
Audited by Gen Agent Trust Hub on Apr 4, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [NO_CODE]: The skill consists exclusively of Markdown documentation and reference files. It does not include any Python scripts, JavaScript files, shell scripts, or binaries.
- [PROMPT_INJECTION]: The skill documents a potential surface for indirect prompt injection within its structured prompting guidance. Evidence Chain: 1. Ingestion points: The file
rules/json-structured-prompting.mdsuggests using template variables such as{{PRODUCT_NAME}}and{{BRAND_HEX_1}}for programmatic prompt generation. 2. Boundary markers: The suggested JSON structures and natural language examples do not include delimiters or instructions to ignore potential commands within interpolated text. 3. Capability inventory: The skill itself possesses no internal capabilities or tools; it provides instructions for interacting with external image generation services. 4. Sanitization: There is no guidance provided for sanitizing, validating, or escaping user-supplied data before it is inserted into the prompt templates.
Audit Metadata