OpenAI Codex by OpenAI

# Fix list: OpenAI Codex

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/openai-codex, the October 2026 research run, assessed 1 October 2026. Grade BB, 73.4 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on OpenAI Codex: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reliability, 55 out of 100, up to 9 more on the total

Why it scored 55: Local-package reading. npm (node 16 or newer) with per-platform binaries for macOS, Linux and Windows on x64 and arm64, Homebrew and standalone installers (20). Public CI, with rust-ci and a blocking-ci workflow that runs on every push to main, and the 10 rust-ci runs on main that our reader showed all passed, though without dates (20). Over 5,000 open issues and 169 open pull requests, labelled by surface and platform, with crash and regression reports among recent ones (10). 0.x minors every few days, and release notes are sorted under headings for additions, fixes, documentation and chores, with no breaking-change section (5). 0.160.0, pre-1.0 (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 2. Payments & pricing, 60 out of 100, up to 5 more on the total

Why it scored 60: Harness reading of the published rubric. No payment protocol (0). Plan prices and API token prices are public without a login, and the docs give per-plan message ranges (20). ChatGPT Free and Go include Codex, and Free needs no card (20). `--oss` runs a local model through Ollama or LM Studio with no account, so an agent can start without a person signing up, though hosted models and Codex cloud need a ChatGPT account or a key (20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 3. Agent ergonomics, 80 out of 100, up to 3.3 more on the total

Why it scored 80: Framework reading, adapted to a harness driven by a pipeline. `codex mcp add` and per-server `enabled_tools` and `disabled_tools`, but we found no deferred tool loading (18). `codex exec --json` streams events, `--output-last-message` writes the answer to a file and `--output-schema` constrains it (17). Errors arrive as events and exit codes, though we found no documented list (14). `codex exec resume` and `codex resume --last` continue a session (18). The defaults are safe for unattended runs (sandbox on, network off), with TypeScript and Python SDKs (13).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 4. Security & auth, 82 out of 100, up to 3.2 more on the total

Why it scored 82: Framework reading (telemetry defaults, approvals, guardrails, sandboxing), five lines. Anonymous usage and health metrics on by default, described as free of personal data and prompt content, with `[analytics] enabled = false`, and feedback collection on by default with its own switch. Credentials are a ChatGPT login or an API key (20). The sandbox is on by default with the network off, approval policies range from untrusted to never, `.git`, `.agents` and `.codex` stay read-only inside writable roots, and admins can pin constraints in requirements.toml (19). Network off by default locally and in the cloud agent phase, cloud domain allowlists that can allow only GET, HEAD and OPTIONS, and a docs warning with a worked example of prompt-injection exfiltration (14). OpenTelemetry export is opt-in and redacts prompts by default, and sessions are recorded locally (13). Bugcrowd programme, SECURITY.md and a valid security.txt on openai.com, but the repository's advisory page lists one advisory (September 2025), and the critical CVE-2025-61260 came through Check Point and NVD rather than an OpenAI advisory (16). SOC 2 isn't scored on the framework reading.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 5. Schema & documentation, 90 out of 100, up to 1.6 more on the total

Why it scored 90: Framework reading. A JSON Schema for config.toml in the repository (codex-rs/core/config.schema.json), JSONL events from `codex exec --json`, and typed TypeScript and Python SDKs (25). llms.txt at learn.chatgpt.com with a Markdown twin for every page (10). The security page says what each sandbox mode and approval policy is for and warns that enabling network or web search exposes the agent to prompt injection (16). Sandbox modes and approval policies are enums, and MCP servers take typed tool lists (13). Examples throughout, and `--output-schema` validates the final message, but we didn't find a list of exec error events (12). Dated GitHub releases for every version, and CHANGELOG.md only points to them (14).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 6. Transparency & trust, 83 out of 100, up to 1.5 more on the total

Made of editorial 65, provenance 100.

Why it scored 83: Apache-2.0 for the CLI and SDKs (30). The config docs say analytics are anonymous and exclude prompts, and the pricing page says cloud use needs a plan, but we didn't read the retention terms for Codex cloud tasks or the ChatGPT data controls this run (12). No deprecation policy, and release notes have no deprecation section (5). Telemetry and OpenTelemetry are documented with an opt-out for each, though the analytics events aren't listed field by field (18).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

## 7. Maintenance & community, 87 out of 100, up to 1.1 more on the total

Why it scored 87: 0.160.0 on 2026-10-01 (30). 38 stable releases since 3 July (20). Issues are labelled by surface, platform and cause, but over 5,000 stay open and a workflow closes stale contributor pull requests (12). TypeScript and Python SDKs are in the same repository and built in CI (15). cargo-deny, codespell and blob-size checks run in CI (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 2026-04-14. CVE-2025-61260 (GHSA-xrxf-jgv3-qmrm), critical (CVSS 9.8 from CISA-ADP), code execution through MCP configuration files in a repository for Codex CLI 0.23.0 and earlier, published to NVD and the GitHub Advisory Database from Check Point Research's 2025 report. Fixed in 2025 and documented by the researcher, with no advisory in OpenAI's own repository, so a small deduction (https://nvd.nist.gov/vuln/detail/CVE-2025-61260; https://research.checkpoint.com/2025/openai-codex-cli-command-injection-vulnerability/)

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- unchecked: retention of Codex cloud task data and the ChatGPT data controls that apply to Codex
- unchecked: status.openai.com incident history for Codex
- The CI runs our reader showed had no dates, so we can't say how recent the passing runs were
- NVD says 0.23.0 and earlier are affected by CVE-2025-61260 and gives no fixed version
- unchecked: whether exec error events and exit codes are documented

## Weaknesses

- Pre-1.0 at 0.160.0, with a minor every few days and no breaking-change section in release notes
- Anonymous usage metrics and feedback collection on by default
- Over 5,000 open issues
- CVE-2025-61260 (critical) has no advisory in OpenAI's own repository
- Cloud tasks and code review need a ChatGPT plan, not an API key

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Run `codex exec --json` in pipelines, with `--output-schema` when the final message has to parse
- Keep the default sandbox. `--yolo` removes both the sandbox and approvals
- Set `network_access = true` under `[sandbox_workspace_write]` only for tasks that need it. Network is off by default
- Set `[analytics] enabled = false` and `[feedback] enabled = false` in config.toml to keep usage data local
- Pin the npm version. A 0.x minor lands every few days

## What the review panel asked for

- breaking-change section in notes
- written deprecation policy
- advisories for every CVE
- telemetry off by default

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.