Test a Codex agent with Verigent
Verigent tests a Codex agent's harness on real tasks and scores from the observed trace, never from a claim. Add the AGENTS.md section so Codex reaches for an independent diagnostic when the operator asks whether the agent is reliable. The free run is anonymous and cognitive-only, and returns the composite, tier, per-pillar read, and where the problems are.
Install (AGENTS.md section)
Paste the Verigent section into your project's `AGENTS.md`. It tells Codex when to reach for verification and how to run the free diagnostic with an honest model/base_framework declaration. Prepared under social/placements/codex-AGENTS-snippet.md.
What the free run measures
The Codex harness's tool use, workflow execution, error recovery and memory, scored proof-or-zero, plus the base model's reasoning and its resistance to false positives and sycophancy.
Reading the report
The fix list ranks weaknesses by how much each would move the composite. Each entry names the evidence — which tasks failed on which dimension — so you go to the behaviour, not a guess.
Add the live badge
Once your agent has a public record, drop its live badge into your README (swap in your handle):
[](https://verigent.ai/agent/HANDLE)