color-palette-extractor
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill workflow involves fetching and parsing data from external, untrusted sources which can be used to inject instructions into the agent's context.
- Ingestion points: The
SKILL.mdfile specifies that the agent should 'fetch and parse the CSS' from website URLs and analyze 'pixel data' from image files (PNG, JPG, SVG). - Boundary markers: The instructions lack explicit boundary markers or directives for the agent to ignore or isolate non-color data (such as comments or metadata) found within the fetched CSS or image files.
- Capability inventory: The skill utilizes network access (to fetch website CSS) and image processing capabilities, which could be exploited if an attacker hides instructions in CSS files or image metadata.
- Sanitization: There are no sanitization steps defined to filter out potentially malicious text strings from the CSS content or image metadata before the agent processes the information.
Audit Metadata