horror-suspense-short-by-1jqwyww
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFENO_CODEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions focus on content generation workflows and prompt engineering best practices. It includes explicit safety guidelines prohibiting unauthorized account changes, asset uploads, or the invention of fake metadata.
- [NO_CODE]: The skill consists entirely of markdown files containing instructions and reference data. It does not include any executable scripts, shell commands, or package dependencies.
- [EXTERNAL_DOWNLOADS]: The skill references external media and case studies hosted on
goodcase.aiandx.com. These links are provided as reference material for the agent to inspect using standard browser or image analysis tools to ensure visual consistency in the generated output. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from a local file (
references/cases.md) to inform its prompt generation logic. - Ingestion points: The agent is instructed to read
references/cases.mdat the start of every invocation. - Boundary markers: The instructions require the agent to write a 'reference contract' and name the specific anchor case used, which helps delimit the influence of the external data.
- Capability inventory: The skill does not define any specialized tools, but instructions suggest using available system tools (browser, image/video viewer) to inspect references. No capabilities for file writing or network exfiltration were found.
- Sanitization: No specific sanitization or filtering logic is defined for the content read from the reference file.
Audit Metadata