Drop the remaining branding reference from the website docs - #14
Conversation
The website README credited another BenchFlow site for the design. Remove it and describe the site on its own terms. The documentation linter already forbids that branding, but only scanned the root docs, docs/ and .agents/ — website/ sat outside the check, which is how the reference survived. Scan the website's markdown too, so this cannot creep back in. Left in place deliberately: the pattern inside the linter itself, which has to name the branding in order to detect it, and the NOTICE attribution, which is an Apache-2.0 obligation rather than branding.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8d8baf130d
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".
| # The public website is documentation too. It sat outside this check long | ||
| # enough to accumulate stale branding the check exists to prevent. | ||
| files.extend(path for path in (ROOT / "website").glob("*.md")) | ||
| files.extend((ROOT / "website" / "src").rglob("*.md")) |
There was a problem hiding this comment.
When stale branding is added to the actual public-site copy, such as website/src/app/page.tsx or website/src/app/contribute/page.tsx, this CI check still succeeds because the new traversal only includes *.md. The current website has no Markdown under src; its user-facing pages are TSX, so this addition protects the developer README but not the rendered website described by the comment. Include the relevant .tsx/.ts source files in the scan so branding cannot return on the public site unnoticed.
Useful? React with 👍 / 👎.
1 participant
Removes the last prose reference to the other BenchFlow benchmark, and makes the existing rule cover the place it survived.
The reference
website/README.mdopened by crediting that project for the site's design. Now it describes the site on its own terms — same information about what the site is, none of the borrowed identity.Why it survived
lint_frontierphysics_docs.pyalready forbids this branding, butdocumentation_files()only collected the root docs,docs/**/*.mdand.agents/**/*.md.website/was never scanned, so the reference sat there passing CI.Now the website's markdown is scanned too — 36 documentation files instead of 35. Verified both directions: re-introducing the phrase fails the check with
ERROR: website/README.md contains stale branding: 'SkillsBench', and it passes once clean.Two mentions left on purpose
lint_frontierphysics_docs.py:20NOTICE:6NOTICE needs your call
NOTICErecords that the repository's task-package conventions, authoring guidance, validation scripts and review tooling are adapted from an Apache-2.0-licensed project. Apache-2.0 §4(d) requires derivative works to carry forward the attribution notices from the original's NOTICE file, so if that adaptation is real, this text is a licence obligation and removing it would put the repo out of compliance.I have not touched it. If the adaptation claim is no longer accurate — the tooling has been rewritten since — then it can be removed as simply stale, but that is a factual call about provenance rather than a branding one, and it is yours to make.
Verification
package-lock.jsonexcluded — dependency metadata).ruffclean on the changed script.