openai / openai/codex-plugin-cc
[BUG] adversarial-review / review fail with HTTP 400 ('gpt-5.5 requires newer Codex') on CLI 0.130.0
Nobody has claimed this yet.
- Dominant language
- JavaScript
- Stars
- 33.3k
- Forks
- 2.3k
- PR merge metrics
- No merged PRs in 30d
Description
Symptom
/codex:adversarial-review (and presumably /codex:review) fails immediately with:
[codex] Codex error: {"type":"error","status":400,"error":{"type":"invalid_request_error",
"message":"The 'gpt-5.5' model requires a newer version of Codex. Please upgrade to the latest app or CLI and try again."}}
[codex] Turn failed.
Result file shows: Codex did not return valid structured JSON. — Parse error: <same 400 above>.
Environment
- Codex CLI:
0.120.0→ upgraded to0.130.0(latest as of 2026-05-09). Same failure on both. - Plugin:
codex@openai-codexv1.0.4 (latest, reinstalled cleanly to confirm). - Auth: ChatGPT login active (verified via
node scripts/codex-companion.mjs setup --json→ready: true,auth.loggedIn: true). ~/.codex/config.toml:model = "gpt-5.5"(default after CLI upgrade).- Direct
codex review --base <ref>works correctly with same model and same config — the issue is plugin-specific.
Reproduction
- Codex CLI 0.130.0, ChatGPT login,
model = \"gpt-5.5\"in config.toml - Run
/codex:adversarial-review --base <ref> --background(or--wait) - Plugin task-thread returns HTTP 400 from Codex API
Root Cause Hypothesis
scripts/codex-companion.mjs adversarial-review (and review) only accept `[--wait|--background|--base|--scope] [focus]` — no `--model` flag (only `task` subcommand accepts `--model`). The companion script falls back to the config.toml default model, which on a freshly-upgraded CLI is `gpt-5.5`. The plugin's task-thread protocol layer apparently doesn't advertise sufficient version to the API for gpt-5.5 dispatch — direct `codex review --base ` (CLI-native, not task-thread) succeeds with the same model and config.
Workaround
Edit `~/.codex/config.toml`: `model = "gpt-5.4"` (older but supported by plugin's task-thread protocol). Revert after invocation.
Suggested Fix
Either:
- Expose `--model ` flag on `review` and `adversarial-review` subcommands (mirror what `task` already does), so users can pin a known-good model per-invocation, or
- Update plugin's task-thread protocol layer to advertise the current Codex version to the API, matching what direct `codex review` does.
Plugin 1.0.4 was published 2026-04-18; gpt-5.5 launched 2026-04-23 — plugin is 5 days behind on protocol compatibility.
Impact
Plugin currently unusable on default-config CLI 0.130.0. New users hitting this immediately after install will get the cryptic "upgrade your CLI" message even though CLI is already at latest.
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start in scripts/codex-companion.mjs, focusing on the review and adversarial-review subcommands and how task handles --model. Reproduce the HTTP 400 with the documented CLI configuration, then compare it with direct codex review --base. Done means the plugin can invoke review with gpt-5.5 successfully or lets users pin the supported model per invocation without the task-thread error.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- javascript, node.js
- Domain
- cli, developer-experience
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 68/100