launch-video
Warn
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from the repository it is analyzing.
- Ingestion points: The workflow requires reading the project's README, documentation, and benchmarks (SKILL.md, Workflow step 1).
- Boundary markers: There are no defined delimiters or instructions to ignore embedded commands within the ingested project files.
- Capability inventory: The skill has access to code synthesis engines, video renderers, file system commands (reveal in Finder), and git operations (commit to repo).
- Sanitization: While it instructs the agent to verify claims, there is no mechanism to sanitize the documentation content to prevent instructions in those files from hijacking the agent's logic.
- [DYNAMIC_EXECUTION]: The skill explicitly calls for the generation and execution of code to handle complex multimedia tasks.
- Evidence: Workflow step 3 describes "Synthesizing the music in code" and step 5 involves rendering high-quality video (H.264/AAC). The mention of "Remotion's manual advance" points to the execution of JavaScript-based motion graphics code.
- [COMMAND_EXECUTION]: The skill performs operations that interact directly with the user's environment and repository.
- Evidence: Instructions include using system commands to "reveal it in Finder" and "commit the project to the repo as a standalone package."
Audit Metadata