sales-pitch-training
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to connect to an external Model Context Protocol (MCP) server at
https://connect.clueso.io/mcp. This resource is owned by the skill's author, clueso-ai, and provides the necessary infrastructure for the video generation workflow. - [COMMAND_EXECUTION]: The instructions include commands to configure the agent's environment, such as
claude mcp add --transport http Clueso https://connect.clueso.io/mcp. This is a standard procedure for adding specialized tools to the assistant's context. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, untrusted data to generate scripts.
- Ingestion points: User-supplied talking points, objection lists, and battlecards provided via the chat interface.
- Boundary markers: The instructions do not define specific text delimiters for the injected content, but they establish a clear workflow boundary.
- Capability inventory: The skill can draft scripts, generate voiceovers, create video scenes, and export files through the connected Clueso tools.
- Sanitization: The workflow incorporates a mandatory human-in-the-loop review step ("show it to the user before building anything"), requiring the user to approve the generated script before the agent performs any further automated actions.
Audit Metadata