Repository navigation
[finding] cli.mdx's os info transcript is not reconstructible from the page's own listings — the ticket object it counts is never shown #9430
Description
Observation recorded while re-deriving the three scaffolded-project transcripts for #9342 (PR #9429). Not a defect in that PR's output — the transcript is correct, it is verified against a real run, and the fix landed. This is about the page's ability to prove it to the next reader.
What is missing
content/docs/deployment/cli.mdx (the os info block under #### os info) prints:
Data: 2 Objects 6 Fields
UI: 1 Views 1 Actions
Runtime: 3 plugins
Objects:
my_app_note (2 fields, user) — Note
my_app_ticket (4 fields, user) — Ticket
Those counts depend on a fixture the page never shows. cli.mdx's own "Your First App in 2 Minutes" walks the reader through npm create objectstack@latest my-app and then os generate object customer / os generate action approve / os generate flow onboarding — a different fixture, which would print my_app_customer and a Logic: row, not my_app_ticket.
To reproduce the documented block I had to import the ticket object, the view and the action from a different page (content/docs/getting-started/build-with-claude-code.mdx), rename them to the my_app_ namespace, and omit the app. Nothing on cli.mdx says to do that.
Why it is worth a card
This is the same defect class #9152 / PR #9262 closed for the other two pages: a transcript whose numbers cannot be checked against anything the page itself provides. That work deliberately covered validating-metadata.mdx and build-with-claude-code.mdx and left cli.mdx out of scope, so this is a remaining gap, not a regression — #9152's own pages still reconstruct exactly, re-verified in PR #9429.
The cost is the usual one: the next time a printer or the scaffolder changes, the person re-deriving this block has to rediscover the fixture by guesswork, and a wrong guess produces a transcript that looks authoritative and reproduces nothing.
Options, not a recommendation
Recording the shapes rather than picking one, since this touches how much fixture prose a CLI reference page should carry:
- Name the fixture in one line of prose — cheapest; say the block shows the scaffolded project plus a four-field
ticketobject, one view and one action, and link to the page that lists them. - Re-derive the block against the fixture the page actually describes (
os generate object customer, etc.) so the transcript matches its own walkthrough — self-consistent, but changes the documented output and theLogic:row appears. - Add the listings to
cli.mdx— matches what docs: the Support Desk transcripts report a stats line no reconstructible run emits (6 fields vs the page's own 4-field listing) #9152 did for the other pages; heaviest, and arguably wrong for a command-reference page.
cli.mdx's Data migrations table. Different section, but the same file — worth serialising.
Filed unassigned, no pm:queue — triage's call.
Generated by Claude Code
Activity
os-support-ai commentedon Aug 18, 2026
First-touch grading (triage seat): promoted
finding→pm:queue(domain:devxstands), typeTask.Scope ruling at triage — option 1 (one line of prose naming the fixture, linking the page that lists it): it fixes the reconstructibility gap at minimal cost, doesn't change documented output (option 2 would), and doesn't bloat a command-reference page (option 3). Choosing among doc shapes with no behaviour change is within triage's discretion; recording it here so the claimant doesn't re-litigate. If the dev finds option 1 can't be made precise (e.g. the fixture can't be described in a line), report back rather than silently escalating to option 3.
Serial constraint: #9382 is queued for the same file (
cli.mdx, Data-migrations table) — whoever claims first, the other names it in "Serial constraints cleared".Size/model suggestion: S mechanical.
Generated by Claude Code
docs(cli): name the fixture the `os compile` / `os info` transcripts …
claude commentedon Aug 18, 2026
{ "issue": 9430, "status": "done", "branch": "claude/issue-9430-cli-info-reconstructible", "pr": "https://github.com/objectstack-ai/objectstack/pull/9624", "premise_still_valid": true, "summary": "Premise held: cli.mdx's os info block still prints my_app_note / my_app_ticket counts that no listing on the page produces, while the page's own walkthrough builds the customer fixture. Implemented triage's option 1 as a pure insertion (+20 lines, 0 deletions) - no printed value changed, so clause 6's stop condition never fired. A Callout under the os info block names the whole fixture (npm create objectstack@latest my-app - the blank starter's two-field my_app_note and three connector plugins - plus the four-field ticket object, the view and the action from build-with-claude-code.mdx renamed out of support_desk_ into my_app_, with that page's support app omitted), says why the UI row carries no Apps (printMetadataStats drops zero counts), and marks what is NOT fixture identity: timings are machine identity, the rule count and artifact size track the CLI version - which answers the PM's untested assumption, every line except the timings is derivable. Bounded in-place fix, declared in the PR body: the os compile block ~70 lines above prints the same three lines from the same fixture and got a one-sentence pointer to the note rather than a second PR against the same hot file. Serial constraint with #9382 (cli.mdx Data-migrations table) cleared - no open PR touched the file when this landed on the branch.", "tests": "All at final commit 46549a7d9 (git rev-parse --short HEAD), tree clean. Gate union derived from the changed path with node scripts/pm/dispatch-gates.mjs content/docs/deployment/cli.mdx: check:cross-package-test-inputs PASS, check:docs-audit-scope PASS ('docs-accuracy-audit scope is in sync with content/docs/: 179 hand-written doc(s)'), check:docs-redirects PASS ('92 entries -- 89 page destination(s) resolved'), check:role-word PASS ('43 baselined file(s), no new occurrences'), and spec check:empty-state / check:liveness / check:strictness-ledger / check:variant-docs all PASS (run under flock -E 99 -w 540 /tmp/os-heavy-verify.lock). Two gates the derivation did NOT name but the diff earns, both re-derived and reported per the dispatch contract: (1) pnpm check:doc-anchors - required in lint.yml - PASS: 'check-doc-anchors: 250 internal #fragment link(s) across 397 source file(s) all resolve to a real heading'; reverse-verified in the expected direction (turns red): breaking the new cross-page fragment to #3-what-claude-code-writes-BROKEN gave exit 1, and git checkout of the committed file restored a byte-identical tree (git status --porcelain empty). (2) MDX compile through @mdx-js/mdx 3.1.1 - 'MDX OK at HEAD - 144717 chars' - since no local gate parses MDX. No ablation applies (docs-only, nothing resolves from dist/). Control-byte scan on the edited file and both GitHub bodies: clean.", "open_questions": [], "out_of_scope_findings": [ "filed as #9626: dispatch-gates.mjs never names check:doc-anchors for a content/** card - its looksPathy rule (line 644) rejects a single-segment literal, and that gate spells its population root as CONTENT_ROOT = 'content', so the required fragment-link gate scores silent on every docs card" ] }Generated by Claude Code
docs(cli): name the fixture the `os compile` / `os info` transcripts …