ci: address MiniMax scout review findings

[Th0rgal]

Summary

• Address advisory OCR findings from ci: default large Lean OCR scout to MiniMax #2138.
• Document the OCR_LLM_TOKEN scout-key fallback.
• Replace a dead reviewer-model test assertion with a real workflow assertion that OCR still uses reviewer.
• Trim MiniMax thinking output before fenced JSON stripping.

Validation

• node --check .github/scripts/ocr-router.js
• node --check .github/scripts/post-ocr-review.js
• node --check .github/scripts/test-ocr-routing.js
• node .github/scripts/test-ocr-routing.js
• bash -n .github/scripts/ocr-git-wrapper.sh
• python3 -m json.tool .github/ocr/rules.json
• PyYAML parse .github/workflows/ocr-review.yml
• actionlint .github/workflows/ocr-review.yml

---

Note

Low Risk
CI/docs and small parsing/test tweaks only; no auth or review-gate behavior changes beyond clearer scout credential docs and safer JSON parsing.

Overview
Follow-up to advisory OCR feedback: scout key docs, MiniMax JSON parsing, and a stronger routing test.

The OCR README now states that OCR_SCOUT_LLM_KEY falls back to OCR_LLM_KEY, then OCR_LLM_TOKEN, matching how the router resolves scout credentials when a dedicated scout secret is not set.

In stripThinking, content is trimmed after stripping MiniMax <mm:think> / <think> blocks and before removing fenced json wrappers, so scout responses with trailing whitespace or newlines still parse reliably.

testScoutConfigDefaultsToMiniMaxOnOcrEndpoint no longer uses a no-op reviewer-model assertion; it reads ocr-review.yml and checks that full OCR still sets OCR_LLM_MODEL: reviewer while scout config defaults to MiniMax on the shared endpoint.

^{Reviewed by Cursor Bugbot for commit d7e7977. Bugbot is set up for automated code reviews on this repo. Configure here.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
verity	Ready	Preview, Comment	Jul 9, 2026 8:28pm

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_e046ae25-bfba-420b-a540-e1165d9a8d6f)

[Th0rgal]

@codex review

[github-actions]

OpenCodeReview first-pass review

Status: success · Mode: config-docs · 0 finding(s) · 3 files · 40299 tokens · 8 tool calls

No comments generated. Looks good to me.

OCR pilot metrics

• Routing: config-docs (router-v5)
• Changed files: 3 supported / 3 total; Lean 0, trust docs 0, workflow/scripts 3, contracts 0, docs 0
• Changed lines: 7 supported; thresholds large Lean >=3 files or >800 lines
• OCR: status success; comments 0; files 3; tokens 40299; tool calls 8; warnings 0; duration 33s
• Largest changed files: .github/scripts/test-ocr-routing.js (+2/-2), .github/ocr/README.md (+1/-1), .github/scripts/ocr-router.js (+1/-0)

Pilot mode: advisory only. Codex Review remains the merge gate.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Already looking forward to the next diff.

Reviewed commit: d7e797726f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".