d6e-workspace-api-client
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill documents an architecture that processes data from various ingestion points (file storage, Markdown documents, and SaaS API proxies) which are interpolated into AI agent prompts via synchronous and asynchronous intent jobs.
- Ingestion points: Data enters the agent context through file uploads (
src/routes/api/upload/+server.ts), workspace documents (references/documents.md), and mirrored Google Drive files (references/drive-sync.md). - Boundary markers: The documentation in
references/chat-streaming.mddescribes a context assembly pipeline that combines prompt rules and system instructions to establish operating boundaries. - Capability inventory: The system possesses capabilities for SQL execution (
d6e_sql), external API calls (d6e_call_external_api), and runtime execution of State Transition Functions (STFs) in JavaScript, Docker, or Wasm environments. - Sanitization: The documentation specifies a sanitization chain that validates tool inputs and filters non-persisted message parts to prevent prompt integrity issues.
- [EXTERNAL_DOWNLOADS]: The skill references canonical source code and implementation examples hosted on the official vendor repositories for d6e-ai.
- Evidence: Documentation includes links to
gitlab.com/cauchye/d6e-ai/d6eandgithub.com/d6e-ai/d6efor reference implementations of the d6e-cloud client and environment configuration.
Audit Metadata