wds-5-agentic-development
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external project specifications and user feedback to drive the development process. This creates a surface for indirect prompt injection where maliciously crafted specifications could attempt to override agent instructions or inject malicious payloads into generated code. Ingestion points: External scenario specifications and user-provided feedback within implementation loops. Boundary markers: The agent follows structured step files with 'Universal Rules' and specific goal definitions to maintain role boundaries. Capability inventory: File system operations (read/write), directory management, and browser automation (Puppeteer). Sanitization: No explicit sanitization or escaping of ingested specification content is documented.
- [DYNAMIC_EXECUTION]: The workflow relies on generating code from templates and executing it via Puppeteer for self-verification. Evidence: Steps in 4d-present-for-testing.md and the testing guide explicitly instruct the agent to run the generated prototype in Puppeteer and verify DOM elements and styles.
- [COMMAND_EXECUTION]: The agent performs automated environment setup, including the recursive creation of project directories and configuration files. Evidence: Step 1 instructions in 1-prototype-setup.md include protocols for creating a multi-directory folder structure and initializing JSON/YAML configuration files.
- [EXTERNAL_DOWNLOADS]: Prototype templates utilize styling and font resources from well-known external services. Evidence: Templates in page-template.html reference the Tailwind CSS CDN and Google Fonts.
Audit Metadata