quilium-builder
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from the Quilium CMS via MCP tool calls (e.g.,
get-site-settingsin SKILL.md anditem listin references/cli-content.md) and has capabilities to execute shell commands (e.g.,quilium content push,quilium site createin references/cli-content.md) and modify site configuration. While the instructions in references/engine-liquid.md recommend using the| escapefilter and mentionstrictFilters, the processing of untrusted content from the CMS environment into the agent's context constitutes a surface for indirect prompt injection attacks.
Audit Metadata