lov-gen-logo
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project data to derive design concepts, creating a surface for indirect prompt injection. * Ingestion points: Reads package.json, README.md, app.json, and Info.plist in SKILL.md (Steps 0 and 1) to extract project metadata and descriptive content. * Boundary markers: Absent; there are no instructions or delimiters in the SKILL.md workflow instructing the agent to ignore potentially malicious instructions embedded in these external files. * Capability inventory: The skill can execute shell commands via magick, vtracer, and svgo (SKILL.md), and it has file-write and git-commit capabilities. * Sanitization: Absent; content derived from README.md is used directly to influence the prompts for the image generator without filtering.
- [COMMAND_EXECUTION]: The workflow relies on external binaries and scripts for its core image manipulation tasks. * Evidence: SKILL.md and scripts/build_app_icon.sh execute shell commands including magick, vtracer, and svgo.
Audit Metadata