html-brief
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest arbitrary content from the user and transform it into a self-contained HTML file for local viewing (
file://). This creates a vulnerability surface where malicious instructions or scripts embedded in the input data could be included in the generated output. - Ingestion points: The skill takes "content" (summaries, status reports, notes, etc.) as input from the agent's context or user prompts.
- Boundary markers: No explicit boundary markers or instructions to ignore embedded instructions in the source data are provided in
SKILL.md. - Capability inventory: The skill uses file-writing capabilities to save the brief and uses the
SendUserFiletool to deliver the rendered output to the user. - Sanitization: The instructions do not explicitly mandate sanitization or escaping of the input content before interpolating it into the HTML shapes. While the template is purely CSS-based, the inclusion of prose as a "shape" could allow for the injection of
<script>tags or event handlers if the agent simply copies input text.
Audit Metadata