favicon-gen
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a pipeline that processes untrusted external files supplied by users to convert them into brand assets.
- Ingestion points: As described in
references/extraction-methods.md, the agent is guided to ingest user-provided logo files, including SVG, PNG, and JPEG files. - Boundary markers: Absent. No specific boundary protection or prompt instructions exist to warn the agent against executing text directives or reading malicious payloads that could be hidden inside SVG text elements or metadata.
- Capability inventory: Detailed in
SKILL.mdandreferences/format-guide.md, the skill invokes local shell capabilities using ImageMagick (convert) to resize and convert files. - Sanitization: Absent. No filtering, validation, or structural sanitization is performed on user images before they are handled by the agent or passed to external command line utilities.
- [COMMAND_EXECUTION]: The instructions require running command-line tools like ImageMagick (
convert) to handle the file generation process. This creates a reliance on the presence and security state of local binaries when interacting with arbitrary, user-controlled graphical files.
Audit Metadata