full-stack-developer
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONOBFUSCATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for building applications that ingest untrusted data from users, such as AI chatbot prompts and web form inputs.
- Ingestion points: API endpoints defined in 'modules/core-guidance.md' (e.g., 'UserInput' Pydantic model) and frontend chat interfaces.
- Boundary markers: The documentation focuses on infrastructure and code logic; it does not provide specific prompt delimitation instructions for internal LLM calls.
- Capability inventory: The skill leverages 'run_terminal_cmd' and file system access as part of its development and deployment automation workflow.
- Sanitization: The skill explicitly includes code examples demonstrating the use of the 'bleach' library for HTML sanitization in user input validators.
- [COMMAND_EXECUTION]: The documentation references a local PowerShell script 'new-fullstack-app.ps1' used for scaffolding production-ready project structures.
- Evidence: Multiple examples in 'SKILL.md' and 'modules/core-guidance.md' show the usage of '. ew-fullstack-app.ps1' with configuration flags for AI, monitoring, and PWA features.
- [OBFUSCATION]: The skill uses Base64 encoded strings for UI image placeholders ('blurDataURL').
- Evidence: Encoded strings in 'SKILL.md' and 'modules/core-guidance.md' (e.g., '/9j/4...') were analyzed and confirmed to be benign JPEG headers used for image loading states.
Audit Metadata