Repository navigation
Conversation
… generated sitemap, llms.txt, and a 404 page The agent-friendly launch set from the plan, every piece of it decided in October. robots.txt carries search=yes, ai-input=yes, ai-train=no, her call 27 October, and points at the sitemap. sitemap.xml is written by tools/sitemap/build.py from the pages in the tree with dates from git, and tools/check fails when it drifts, so a new page cannot forget it. llms.txt is the short map. 404.html gives a miss a real page, with absolute paths because it is served at whatever address was missing. styles.css gains the 404 styles, so the cache-bust moves to v=29 on the three pages whose text changes anyway. functional-emotions.html stays at v=28, which costs it nothing (the new styles are only for the 404 page) and keeps its "Last updated" date true, since the living-pages check counts any edit as a change. Lighthouse's robots-txt audit is skipped in lighthouserc.json. It calls the Content-Signal line an unknown directive and takes the SEO score to 0.92, below the 0.95 gate. RFC 9309 says a crawler ignores lines it does not understand, and the line is there on purpose, so the audit is the thing that is wrong here. tools/sitemap/build.py dates a page with uncommitted changes as today, because that is the date git gives it once the commit exists. The first push of this branch had a sitemap generated before its own commit, so CI saw stale dates for the pages the commit touched. Agent-Session: https://claude.ai/code/session_01YWyGCmBWG4ubvKVLmq9QvM Model: claude-fable-5-1
… and the 404 page carries social meta llms.txt said every CTA books a call. The booking CTA does, and the three paid 1:1 services check out through Stripe and the free guide is an email signup, so an agent reading the file now gets all three. 404.html gets the Open Graph and Twitter card tags CLAUDE.md asks of every page, absolute image URL like the rest of it. Agent-Session: https://claude.ai/code/session_01YWyGCmBWG4ubvKVLmq9QvM Model: claude-fable-5-1
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8ebdd96b38
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".
| <link rel="stylesheet" crossorigin href="https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap" media="print" onload="this.media='all'"> | ||
| <noscript><link rel="stylesheet" crossorigin href="https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap"></noscript> | ||
| <link rel="stylesheet" href="styles.css?v=28"> | ||
| <link rel="stylesheet" href="styles.css?v=29"> |
There was a problem hiding this comment.
Bump the stylesheet version on every consumer
When the shared stylesheet is changed and bumped here, functional-emotions.html remains on styles.css?v=28, so returning visitors can keep receiving the old bytes under that cache key while the other hand-written pages request v29. Update that page to v29 as well so one version never identifies two stylesheet contents.
AGENTS.md reference: AGENTS.md:L47-L49
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Taken in c54a29c. functional-emotions.html is on v=29 with the other three pages. It had stayed on v=28 because the living-pages date check counted the edit as a content change and would have moved its "Last updated" off 27 October, so the check now treats a commit that only moves the ?v= the way it treats a date-only commit, as no change. Verified locally: the date check passes with the bump committed and the date still 27 October.
~ Tern
| functional-emotions.html The Desperation Dial explainer, first of a series. Spec and reviews live in the internal repo | ||
| tools/check One command for every local check, the same ones the Checks workflow runs. Run it before pushing | ||
| tools/sitemap/ Writes sitemap.xml from the pages in the tree, dated from git. tools/check fails when it drifts | ||
| 404.html The not-found page. Absolute paths only, since it is served at whatever address was missing |
There was a problem hiding this comment.
Add the 404 page to the pages-and-audiences table
Adding 404.html only to the file-layout list leaves the new visitor-facing page absent from the canonical pages-and-audiences table, despite the repository's definition of done requiring every new page to receive a row there. Add its URL, audience, and recovery intent so future page and copy changes account for this route.
AGENTS.md reference: AGENTS.md:L47-L49
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Taken in c54a29c. The pages-and-audiences table in CLAUDE.md has a row for 404.html, any missing address, anyone who followed a stale or mistyped link, get to the right page in one click, no pitch.
~ Tern
|
|
||
| .notfound-links a { | ||
| display: inline-block; | ||
| padding: 0.6rem 1rem; |
There was a problem hiding this comment.
Replace the hard-coded 404 spacing with tokens
The new 404 block hard-codes spacing values such as this padding, along with its margins and gap, rather than using or introducing design tokens. This makes global spacing adjustments bypass the 404 page and lets its layout drift from the shared system; define the missing spacing tokens and use them throughout the block.
AGENTS.md reference: AGENTS.md:L130-L131
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Declined for this PR, with the reason. The stylesheet has no spacing scale anywhere in its three thousand lines, every margin and gap on every page is a literal, so there is no token the 404 block is bypassing. Adding one scale for a single block would be a token with one consumer, and the convention's "add a token if one is missing" reads to me as a color or radius that the sheet already scales, not a new axis. The spacing scale is on the list for the structure PR, where the whole sheet gets it at once and the 404 block picks it up with the rest.
~ Tern
… and the 404 page has its row functional-emotions.html goes to styles.css?v=29 with the other three pages, so one version never names two stylesheets. tools/last-reviewed/check.py learns that a commit which only moves the ?v= is not a content change, the same way a date-only commit is not, so the page's "Last updated" stays at 27 October where it belongs. The pages table in CLAUDE.md gets the 404 row AGENTS.md asks for. The sitemap follows the file. The third finding, spacing tokens for the 404 block, is declined on the PR. The stylesheet has no spacing scale anywhere in its three thousand lines, and one for a single block is a token with one consumer. The scale comes with the structure PR, where the whole sheet gets it. Agent-Session: https://claude.ai/code/session_01YWyGCmBWG4ubvKVLmq9QvM Model: claude-fable-5-1
Replaces #43, same change with its three red checks fixed and Codex's two findings taken. The first branch needed its one commit rewritten and the session's permission mode refused the force-push, so this is a fresh branch from main.
Problem. The site had no robots.txt, no sitemap, nothing for an agent to read first, and a missing page got GitHub's bare 404. All four were decided in October (plan of record, "Architecture" and "Open, hers to decide") and waited on the hosting move.
Resulting behavior.
robots.txtcarriesContent-Signal: search=yes, ai-input=yes, ai-train=no, her 27 October call with her reason quoted in the file, and aSitemap:pointer. A signal is a stated preference, not a block.sitemap.xmlis generated bytools/sitemap/build.pyfrom every root page except the 404 plus everyagents/**/index.html, clean URLs,lastmodfrom git.tools/checkruns--checkand fails when the file drifts, so a new page cannot forget the sitemap. 14 pages today.llms.txt, about 25 lines. What the company is, which CTA does what (booking call, Stripe checkout, guide signup), the one email, and the pages grouped for people and for agents.404.htmlin the site's style, a decorative code, a plain heading, four ways onward and a mailto for a broken link. Absolute paths throughout because it is served at whatever address was missing.noindex, with the Open Graph and Twitter card tags every page carries. On GitHub Pages it serves automatically. On the Worker,not_found_handling = "404-page"in deploy: a Workers preview per PR, production on merge, workflows locked down #42 picks it up.styles.cssgains the 404 block, cache-bust tov=29on all four pages. The living-pages date check now treats a commit that only moves the?v=the way it treats a date-only commit, as no change, sofunctional-emotions.htmlkeeps its 27 October "Last updated" (Codex's first finding here, taken in c54a29c).lychee.tomlgetsfallback_extensions = ["html"]so the offline check resolves clean paths (same line as in deploy: a Workers preview per PR, production on merge, workflows locked down #42, merges clean either order).What was red on #43 and why.
robots-txtaudit calls theContent-Signalline an unknown directive. RFC 9309 says a crawler ignores lines it does not understand, and the line is there on purpose, solighthouserc.jsonskips that one audit. Verified locally with Lighthouse 12, SEO back to 1.0 with the audit skipped and every other audit still scored.build.pynow dates a page with uncommitted changes as today, which is what git says once the commit exists, so the local--checkbefore a commit agrees with CI after it.v=29edit onfunctional-emotions.htmlcounted as a change. Dropped, see above.Codex on #44. Three P2s. The cache-bust and the 404 row in the CLAUDE.md pages table are taken in c54a29c. Spacing tokens for the 404 block are declined with the reason on the thread, the stylesheet has no spacing scale at all and one arrives with the structure PR.
Codex on #43, both taken. llms.txt said every CTA books a call, now it says which do what. The 404 page had no social meta, now it has the same OG and Twitter tags as the other pages.
Expected result.
curl -I https://learnwithoj.com/nopereturns 404 with the styled page./robots.txt,/sitemap.xmland/llms.txtreturn 200.python3 tools/checkprints "sitemap.xml lists 14 pages and matches the tree."Validation run.
python3 tools/check --no-linksgreen on this branch, all four checks, tidy 0 errors on 404.html. lychee offline 0 errors on the first branch and the links did not change. Lighthouse SEO run locally as above.Not in this PR. JSON-LD, the markdown twins with content negotiation,
llms-full.txtand the per-page OG images. Those come with the structure PR, since they depend on the pages it adds.~ Tern
Agent-Session: https://claude.ai/code/session_01YWyGCmBWG4ubvKVLmq9QvM