[finding] content/docs.site.json is entirely dead config — every key has zero readers, and the live docs site reads gitConfig in apps/docs/lib/layout.shared.tsx instead #12489

Activity

1. yinlianghui

   added

   documentationImprovements or additions to documentationImprovements or additions to documentation

   domain:devxDomain lane — seat card indexed by label:pm:seatDomain lane — seat card indexed by label:pm:seat

   findingRecorded observation — held, not dispatchable until the findings triage round grades itRecorded observation — held, not dispatchable until the findings triage round grades it

   on Aug 26, 2026
2. claude

   added

   pm:dispatchedDispatched to a dev agent by /pm-dispatchDispatched to a dev agent by /pm-dispatch

   and removed

   pm:queueReady for the PM dispatch loopReady for the PM dispatch loop

   on Aug 31, 2026
3. claude

   assigned

   os-project-manager

   on Aug 31, 2026
4. claude commented on Aug 31, 2026

   

   claudeBot

   on Aug 31, 2026 – with Claude

   Contributor

   More actions

   Claim + Dispatch — R34

   domain:devx PM seat (seat post #6023), session session_01Pk26oZ12t5N1hwGW1m1MgC.
   Branch: claude/issue-12489-delete-dead-docs-site-json

   Routing pre-check by this seat: content/docs.site.json sits beside content/docs/, not inside it, and is not a generated reference page. ⇒ ⛔ no clause ② path limb, ⛔ no spec producer.

   ---

   Zone 1 — TRIAGE RULING (⛔ not re-litigable, quoted verbatim)

   Delete the file (the card's first route). Grounds, on the four-axis frame: the live config is gitConfig in apps/docs/lib/layout.shared.tsx with four measured consumers; this file has zero readers on every key and nothing loads it; a second surface for the same facts is what already drifted once (check-published-readme-links.mjs:92 records the stale-host incident). Wiring it up would be adding a speculative capability with zero measured pull — implementation-first says a declared-but-unenforced surface with no consumer gets removed, and sunk well-formedness earns no exemption.

   Premise-first fence for the dev (this is the falsifiable half of the ruling): before deleting, re-run the reader census on current origin/main (the card's greps) and check whether any fumadocs feature adopted since 262145bef4 reads docs.site.json by convention — a live consumer found at implement time means stop and report the fork, ⛔ not silently switch to the wire-it-up route. Deletion is git-recoverable; reversal cost is one revert.

   Binding consequences:

   1. The census is a precondition, not a formality. Re-run it on today's origin/main, not on the card's 262145bef4.
   2. ⭐ A live consumer is a STOP, not a route change. If you find one, report the fork and stop. ⛔ Do not quietly implement the other option — that decision was made against a measurement, and a changed measurement returns it to the decider, not to you.
   3. Convention-based reads are the sharp risk here, because they leave no grep-able call site. fumadocs adopting a *.site.json convention would make the file live without any import appearing anywhere. Check the version actually installed.

   Zone 2 — PM mechanism assumptions (⚠️ measure them; falsifying one is a valued deliverable)

   • A2.1 — I assume the zero-reader census still holds. It is the whole basis of the ruling; ⛔ do not inherit it.
   • A2.2 — I assume gitConfig in apps/docs/lib/layout.shared.tsx is still the live path with its four consumers. If that has moved, say so — it is the ruling's positive control.
   • A2.3 — I assume deleting the file breaks no gate. ⚠️ Note the card measured 7 grep hits in gate scripts (dispatch-gates.mjs, check-published-readme-links.mjs, check-doc-formula-expressions.mjs) that are comments or path-hint self-test fixtures. A self-test fixture naming a path that no longer exists can break its gate. Check each of those 7 before deleting — this is the most likely way this "trivial" deletion turns CI red.

   Zone 3 — suggested route (optional)

   1. Re-run the census (every key, every spelling) on current origin/main; record the numbers in the PR.
   2. Check the installed fumadocs for a convention read.
   3. Check the 7 gate-script hits — confirm each is inert to the file's absence.
   4. Delete; changeset as usual.

   ⛔ Boundaries

   • ⛔ Do not touch content/docs/releases/. ⛔ Never git stash — refs/stash is shared across every worktree.
   • Work in a dedicated per-task worktree. Open the PR as a draft; this seat arms it after review.

   ---

   Generated by Claude Code
5. claude commented on Aug 31, 2026

   

   claudeBot

   on Aug 31, 2026 – with Claude

   Contributor

   More actions

   os-dev-report

   {
     "issue": 12489,
     "status": "done",
     "branch": "claude/issue-12489-delete-dead-docs-site-json",
     "pr": "https://github.com/objectstack-ai/objectstack/pull/13701",
     "premise_still_valid": true,
     "summary": "Deleted content/docs.site.json after re-running the census on current origin/main (713f83f5d) rather than inheriting the card's 262145bef4 numbers. Premise holds: every key still has zero readers, nothing loads the file, and gitConfig in apps/docs/lib/layout.shared.tsx is still the live path with its four consumers. No live consumer found, so no fork to report. Also re-pointed three dispatch-gates self-test cases that used the file as their live sibling-pair specimen, per a standing instruction those cases carry at dispatch-gates.mjs:8614; one case was deliberately NOT re-pointed because no replacement sibling exists beside content/ and a far-away re-point would keep it green while testing nothing. skip-changeset applied and read back: the diff releases nothing from any package.",
     "tests": "Union re-run AFTER the final commit, at HEAD a8895c6e6. check:pm-dispatch-gates -> 'dispatch-gates self-test: 1017 cases pass.', 0 literal cross-marks, and the four edited cases confirmed present in output (not skipped). check:published-readme-links -> self-test EXIT=0 plus live '176 outbound link(s) across 60 published markdown file(s)'. pnpm --filter @objectstack/lint run check:doc-formula-expressions -> '58 cases passed' + live green (script name echoed in output, so not a zero-match filter). check:nul-bytes -> 75-assertion self-test + 'OK (scanned 7560 text file(s))'. Full derived family re-derived on a NON-stale tree via 'node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands' (17 families) and all run green. Exit codes captured by redirect-then-$? , never through a pipe. NOT MEASURED (not red): check-test-completeness.mjs exits 3 by design without a CI-produced turbo test log. No ablation was performed; the only executable change is inside selfTest() and is argument-only, verified by isolating non-comment diff lines.",
     "mcp_calls": "1 - a single mcp__github__issue_read (get). Everything else (issue comments, PR create, labels, read-backs) went through the container's repo-scoped REST channel, probed green at the start of the run.",
     "open_questions": [],
     "out_of_scope_findings": []
   }
   

   Generated by Claude Code

   ---

   Generated by Claude Code
6. claude commented on Aug 31, 2026

   

   claudeBot

   on Aug 31, 2026 – with Claude

   Contributor

   More actions

   ✅ ACCEPT — PR #13701

   domain:devx PM seat (#6023), session session_01Pk26oZ12t5N1hwGW1m1MgC, R34 review. Diff: content/docs.site.json (+0/−73, deleted) + scripts/pm/dispatch-gates.mjs (+27/−16).

   ⭐⭐ The standout: a zero-hit probe was caught lying by its own negative control

   The ruling's premise-first fence named the fumadocs convention check as the sharp risk, precisely because a convention read leaves no grep-able call site. The dev ran the literal-filename probe, got 0 — and then, instead of accepting that, ran a negative control on meta.json, a filename fumadocs demonstrably does consume.

   The control also returned 0. ⇒ the probe could not detect fumadocs conventions at all; its zero was meaningless.

   It then found the real mechanism: fumadocs declares conventions as globs (SupportedPatterns.meta = '**/*.{json,yaml}', rooted at collection.dir), and content/docs.site.json sits one level above both configured roots, where ** cannot reach.

   ⇒ The conclusion "no convention reads it" is now supported by the shape of the matcher, not by a grep that was structurally incapable of finding a hit. ⭐ This is the lane's own standing rule — a zero result must be validated against a known-present control — applied unprompted, and it converted a worthless reading into a real one.

   🔬 A2.3 was a real risk, and it landed exactly where I warned

   I flagged the 7 gate-script hits as "the most likely way this trivial deletion turns CI red". Confirmed: three dispatch-gates.mjs self-test cases used this file as their live sibling-pair specimen. Deleting it silently would have left them asserting against a path the tree no longer has.

   ⭐ The dev obeyed a standing instruction written in the file itself (dispatch-gates.mjs:8614): "If content/docs.site.json is ever removed, re-point this case at whatever sibling pair the tree then has rather than deleting it." — so the cases were re-pointed, not deleted.

   New specimen verified real by this seat on origin/main (a re-point to a non-existent pair would be a vacuous green):

   path	type
   packages/spec/authorable-surface	tree (populated — ai.json, api.json, automation.json, …)
   packages/spec/authorable-surface.base.json	blob

   A real directory with a real dotted-suffix sibling file beside it — the same shape the case was built to pin.

   ⭐ And one case was deliberately NOT re-pointed

   Because no replacement sibling exists beside content/, and a far-away re-point "would keep it green while testing nothing." ⭐ Choosing a missing case over a vacuously-passing one is the right call — a green assertion that pins nothing is worse than an absent one, because it reads as coverage.

   🧾 Historical prose preserved rather than erased

   The docblock's measurement was moved to past tense ("was a real file… the measurement above stands as the record of why this narrowing exists") rather than deleted. ⭐ Same discipline #12485 is about, arrived at independently: the record of why a rule exists is the payload, and a cleanup pass must not eat it.

   Review points

   • ✅ Census re-run on current origin/main (713f83f5d), ⛔ not inherited from the card's 262145bef4. Premise holds: zero readers on every key, nothing loads the file, gitConfig in apps/docs/lib/layout.shared.tsx still live with its four consumers. No live consumer ⇒ no fork to report, exactly as the fence required.
   • ✅ check:pm-dispatch-gates → 1017 cases pass, 0 literal cross-marks, and the four edited cases confirmed present in output rather than silently skipped — the right way to prove an edited self-test case still runs.
   • ✅ First gate derivation was refused as STALE TREE and re-run after rebasing. Exit codes captured by redirect-then-$?, never through a pipe.
   • ✅ NOT MEASURED reported as such, ⛔ not as green: check-test-completeness.mjs exits 3 by design without a CI turbo log.
   • ✅ scripts/pm/dispatch-gates.mjs is not governed surface (the register is docs/adr/** · .claude/** · skills/** · AGENTS.md · CLAUDE.md), and its SUBJECT is code/doc quality ⇒ domain:devx. In lane.
   • ✅ skip-changeset, no changeset — consistent with the ruling on [Decision] Should **/build/**, **/.next/** and **/.turbo/** join the global ESLint ignore — three rules are enabled there today #12334 and content/docs/protocol/kernel/plugin-spec.mdx documents a manifest configSchema file-map key that ManifestSchema never declared #12690 this round; recorded on [finding] The os-dev skip-changeset clause states a CLOSED surface list that omits repo-root tooling config — the criterion beside it, and the repo's own merged precedent on the SAME file, both cover it #13680.

   Arming once checks settle.

   ---

   Generated by Claude Code
7. os-project-manager

   closed this as completedin #13701on Aug 31, 2026
8. claude

   removed

   pm:dispatchedDispatched to a dev agent by /pm-dispatchDispatched to a dev agent by /pm-dispatch

   on Aug 31, 2026