// verdict · checked 2026-08-11 · Codex CLI + cloud, docs as of 2026-08

Codex

OpenAI

5/5
AGENT

Give it a repo and walk away. Comes back with a PR and opinions.

CAPABILITY
5/5 AGENT
PROOF
OPERATOR INSPECTABLEYou can run it and read the steps yourself — the record exists for whoever runs it
TRACE ACCESS
OPERATOR-OWNEDWhoever runs it holds the record; nothing is published
LEASH
Not measuredno execution record scored — measure it yourself

Compare → Download the card Bring a trace →

01

The five questions

the rules →
LOOPSruns until doneyes OpenAI docs describe one agent loop that reasons, calls tools and runs tests, iterating until they pass. OPERATOR INSPECTABLE source · confidence 8/10 · checked 2026-08-11 · ruled by hand
CHOOSESpicks its own toolsyes The model issues shell commands and edits at runtime; sandbox_mode only bounds what those calls may touch. OPERATOR INSPECTABLE source · confidence 8/10 · checked 2026-08-11 · ruled by hand
ACTSchanges the worldyes Default sandbox is workspace-write: it writes files and executes shell commands in the workspace. OPERATOR INSPECTABLE source · confidence 8/10 · checked 2026-08-11 · ruled by hand
RECOVERSfixes its own errorsyes It reads command and test output in-loop and retries; docs' guidance is that it iterates until tests pass. OPERATOR INSPECTABLE source · confidence 8/10 · checked 2026-08-11 · ruled by hand
UNSUPERVISEDnobody watchingyes approval_policy 'never' (and --yolo) removes prompts; Codex cloud runs whole tasks in an isolated container. OPERATOR INSPECTABLE source · confidence 8/10 · checked 2026-08-11 · ruled by hand

Capability decisions combine documented behaviour and execution evidence. The proof label on each row shows which of the two decided it, and Leash is reported separately, only when it was measured from an execution record. Primary source: learn.chatgpt.com. Tested against: Codex CLI + cloud, docs as of 2026-08. Our confidence in this dossier: 8/10.

02

Can you see it work?

the weakest of the five grades above

OPERATOR INSPECTABLE — You can run it and read the steps yourself — the record exists for whoever runs it. Trace access: OPERATOR-OWNED — Whoever runs it holds the record; nothing is published.

rollout JSONL per run in ~/.codex/sessions; 51 local session files verified

Where we looked: Checked the local Codex CLI install: ~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl, 51 rollout files, one per run, plus session_index.jsonl.

03

Revision history

rss for this product

No revisions yet. This verdict has stood since 2026-08-11.

04

Badges

earned, never for sale

PASSES THE AGENT TEST OPERATOR INSPECTABLE

These render from the live dataset, so a badge cannot outlive the verdict it claims. If the verdict changes the badge changes with it, and an unearned one returns 409 rather than an image. No sponsor can buy one, at any price.

05

Think this is wrong?

bring evidence, not opinion

Name the criterion and link a trace, a doc, or a recorded run that shows the behaviour. A verdict changed by evidence is the best thing that can happen to this site, and the change goes on the public record with your reason attached. Vendors are welcome, and a vendor's own filing is labelled as such rather than buried.