twist-comedy-skit
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill workflow in
SKILL.mdand the evidence inreferences/cases.mdinclude URLs pointing togoodcase.ai,media.goodcase.ai, andx.com. The agent is instructed to 'Inspect the anchor's finished media' using available tools to ensure visual fidelity during the prompting process. - [INDIRECT_PROMPT_INJECTION]: The skill processes third-party prompt excerpts and summaries provided in the
references/cases.mdfile. This data is used to derive structure and traits for new content. - Ingestion points: The agent reads content from
references/cases.md, specifically the 'Summary' and 'Prompt excerpt' sections which contain text from external creators. - Boundary markers: The skill requires the creation of a 'Reference contract' (Preserve/Replace/Avoid) to manage the influence of external data on the final output.
- Capability inventory: The agent uses file-reading and network-access (browsing/inspecting) capabilities across the workflow.
- Sanitization: No explicit sanitization, filtering, or 'ignore instructions' warnings are applied to the ingested prompt excerpts.
Audit Metadata