This document describes the AI agents used in the Deaf-First Platform ecosystem, their capabilities, constraints, and integration points.
The Deaf-First Platform uses AI agents to automate workflows, assist users with accessibility needs, and manage CI/CD processes.
- Purpose: Automates code changes, CI fixes, PR creation, and repository maintenance.
- Capabilities: Code generation, file editing, branch management, PR creation.
- Constraints: Cannot push directly; all changes go through GitHub pull requests. Follows repository coding guidelines.
- Purpose: Provides AI-powered features for the Deaf-First SaaS platform including sign language recognition and accessibility workflows.
- Capabilities: OpenAI API integration, custom model endpoints.
- Constraints: API keys must be stored as environment secrets. No hardcoded credentials.
- Purpose: Biometric and accessibility-first authentication for deaf and hard-of-hearing users.
- Capabilities: Sign language recognition for authentication, accessible MFA.
- Constraints: All biometric data must be processed locally; no PII sent to third parties.
- Purpose: Real-time sign language video synchronization and captioning.
- Capabilities: Video stream processing, live caption generation.
- Constraints: Video data is not stored beyond session scope.
- All agents run with the principle of least privilege.
- No agent may commit secrets, credentials, or private keys.
- Agent activity is audited via GitHub Actions logs.
For agent-related security concerns, contact: security@mbtq.dev