brack — agentic threat model
Brack is a lightweight, low-risk security middleware designed to filter malicious inputs rather than act as an autonomous agent. Its primary risk lies in potential bypasses of its regex and Gemma3-based filtering mechanisms, which could expose downstream agents to prompt injection.
OWASP AIVSS score rationale
| Autonomy of Action | 0.10 | |
| Goal-Driven Planning | 0.00 | |
| Self-Modification | 0.00 | |
| Dynamic Tool Use | 0.10 | |
| Persistent Memory | 0.00 | |
| Contextual Awareness | 0.20 | |
| Dynamic Identity | 0.00 | |
| Multi-Agent Interactions | 0.10 | |
| Non-Determinism | 0.30 | |
| Opacity & Reflexivity | 0.20 |
Scored with the canonical OWASP AIVSS formula (AIVSS calculator reference); agentic risk factors estimated from the agent’s described capabilities.
MAESTRO 7-layer threat model
Per-layer threats for this agent. Layers tagged “not certain from listing” are general, caveated commentary where the public description didn’t pin that layer.
Employs a lightweight Gemma3 270M model for intent checks. Threats include adversarial prompt injections designed to bypass this specific model's classification boundaries, or model evasion techniques that exploit the small parameter size of Gemma3.
Not certain from the listing — Brack acts as an inline input filter and does not explicitly mention maintaining its own vector stores, RAG databases, or training data pipelines.
Not certain from the listing — Brack is designed to protect agent frameworks rather than operating as one itself. It does not disclose internal orchestration, planning, or tool-calling mechanisms.
Not certain from the listing — The deployment architecture (e.g., sidecar, API gateway, or library integration) and associated sandboxing or secrets management controls are not specified.
Strongly aligned with this layer. It provides early-stage filtering, input hygiene, and salted HMAC logging to ensure audit trail integrity and prevent log tampering or blind spots.
Acts as a security and compliance control layer. The use of salted HMAC logging supports compliance audits and data integrity verification, directly addressing policy enforcement and input sanitization requirements.
Not certain from the listing — While designed to protect autonomous agents within an ecosystem from malicious instructions, Brack's own direct multi-agent interactions or trust relationships are not detailed.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).