runway-dev-recipes
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external instructions from the RunwayML documentation (
llms.txt) to determine recipe logic and implementation steps. - Ingestion points: The agent is instructed to read
https://docs.dev.runwayml.com/llms.txtand its linked documents to recommend recipes and follow input schemas (SKILL.md). - Boundary markers: There are no explicit delimiters or specific instructions to treat the remote documentation content as untrusted data.
- Capability inventory: The skill uses MCP tools (
get_credit_balance,get_task) and generates SDK client code for application integration. - Sanitization: The skill does not specify any sanitization or validation of the content retrieved from the remote documentation before using it to influence agent behavior.
- [EXTERNAL_DOWNLOADS]: The skill fetches documentation and technical specifications from the official RunwayML developer portal (
docs.dev.runwayml.com).
Audit Metadata