seedance
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (story ideas) to generate complex prompts for video generation models.
- Ingestion points: The workflow in
SKILL.md(Steps 1 and 2) ingest user-supplied loglines and descriptions to create planning documents. - Boundary markers: The skill does not define specific delimiters or warnings to prevent malicious instructions within the user's story ideas from influencing the agent's output during the JSON generation phase.
- Capability inventory: The skill utilizes the
vsCLI vianode dist/cli.js(Step 5) to validate generated files and estimate costs. - Sanitization: No explicit sanitization or escaping of the user's input strings is specified before they are interpolated into the
stills.jsonorshots.jsonfiles. - [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Node.js CLI tool to perform validation and dry-runs of the produced film assets.
- Evidence:
SKILL.mdspecifies commands such asnode dist/cli.js stills films/<slug>/stills.json --dry-run. This execution is part of the intended developer workflow for thevideo-studioenvironment. - [EXTERNAL_DOWNLOADS]: The skill references an external discovery index for a prompt optimization tool hosted by a well-known technology provider.
- Evidence:
references/seedance-prompting.mdmentions fetching a skill index fromhttps://arkdocs-en.tos-ap-southeast-1.volces.com/skills/.well-known/agent-skills/index.json. This URL belongs to BytePlus (ByteDance), an established cloud service provider.
Audit Metadata