documentary-brutalist-agency
Fail
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: CRITICALINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions in
SKILL.mdcreate a data ingestion surface where external, user-supplied content is processed and interpolated into generated source code. - Ingestion points: The agent is instructed to use user-provided photography, projects, quotes, and claims to populate the website layout as described in
SKILL.md. - Boundary markers: There are no instructions providing delimiters or specific guidance to the agent to ignore instructions that may be embedded within the user-supplied project data.
- Capability inventory: The skill generates executable frontend files, including HTML, CSS, and JavaScript, as demonstrated in
demo/index.html. - Sanitization: The instructions focus on accessibility and visual hierarchy but do not explicitly require the agent to sanitize or escape user-provided strings before rendering them into the HTML, which presents a potential risk of cross-site scripting (XSS) in the generated output.
Recommendations
- CRITICAL: 2 file(s) identified as malware by FileRep - DO NOT USE
Audit Metadata