Termyte journal

Technical proof for practical autonomy.

Measured explanations of coding agent guardrails, runtime safety, policy, and engineering workflow.

Workflow / June 4, 2026

Coding Agent Guardrails: How to Let Agents Work More Autonomously

More autonomy does not require blind trust. It requires clear coding agent guardrails and a local decision path.

Read article →
Benchmark / June 4, 2026

Agent Command Safety Benchmark: 1,200 Decisions Measured

How Termyte measures deterministic agent command safety without presenting a fixture suite as proof of complete safety.

Read article →
Runtime / June 4, 2026

Coding Agent Security: What Runtime Mode Limited Means

The agent runtime safety boundary is useful today, but its interception limits must remain explicit.

Read article →
Threat model / June 4, 2026

What Are Agent Guardrails? Why Termyte Is Not a Sandbox

A useful agent guardrail becomes more credible when its non-goals are explicit.

Read article →