descript-operator
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use a local fallback mechanism that executes the
auto-editorcommand via a bash shell when Descript credits are unavailable. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted media recordings and transcripts, which represents a surface for indirect prompt injection attacks.
- Ingestion points: External media files and URLs (Direct, Drive, Dropbox) processed via the
import_mediatool defined inSKILL.md. - Boundary markers: There are no specified delimiters or instructions to ignore embedded commands within the media or generated transcripts.
- Capability inventory: The skill utilizes bash command execution (
auto-editor), file uploads to provided URLs, and network-based updates to a Notion Content Engine. - Sanitization: The instructions do not define any sanitization or validation of the media content or transcripts before they are used to drive agentic operations.
Audit Metadata