TITMAS Health

AI Agent Health Assessment Engine

AI agents are becoming digital workers.
TITMAS Health helps developers and operators understand whether their agents are:

• observable
• stable
• verifiable

This repository is the first-value-entry point for TITMAS:

• run a health check,
• understand the result,
• then connect to deeper runtime and governance components.

It is not a certification system, and it does not provide a safety guarantee. The product position is assessment and measurement, not ranking.

Health Check Dimensions (v0.1)

TITMAS Health v0.1 focuses on three dimensions:

1) Execution Stability

Measures whether an agent performs tasks consistently.

Indicators:

• task_success_rate
• failure_rate
• execution_consistency

2) Evidence Integrity

Measures whether execution history can be verified.

Indicators:

• record_completeness
• hash_verification
• trace_availability

3) Recovery Ability

Measures whether an agent can recover from failure and continue service.

Indicators:

• retry_success
• failure_recovery
• adaptation

Quick Start

Agent Execution
        ↓
TITMAS Health Check
        ↓
Health Report

1. Collect or point to execution artifacts produced by your agent.
2. Generate a health report from the three dimensions.
3. Export the report to a stable schema and review it against repo-specific gates.

Repository Structure

titmas-health/
├── README.md
├── health/
├── metrics/
├── reports/
├── examples/
├── schemas/
├── docs/
└── tests/

Suggested initial paths:

• schemas/health-report.schema.json: canonical output schema for v0.1 reports.
• examples/health-report-v0.1.json: reference report payload.
• docs/health-model.md: model and decision definitions.
• docs/metrics.md: metric definitions, units, and interpretation.
• docs/roadmap.md: release and governance milestones.

Ecosystem Mapping (v0.1)

• Evidence layer inputs: agent-evidence
• Evaluation layer inputs: SAEE
• Accountability layer references: aro-audit

Limitations

Current implementation intentionally avoids:

• certification claims
• intelligence ranking
• safety guarantee claims

TITMAS Health v0.1 is a practical entry point for observability and governance readiness, not final judgment.

Development Guidance

• Keep report outputs explicit and machine-readable.
• Prefer deterministic outputs for reproducible health checks.
• Use the schema in schemas/ before any downstream sharing.