Repository navigation
spec(FIX-1770): name the assertions no-page-config actually fails - #2774
Conversation
The control removes only the page config. TEAMS comes from the page's API reads, so it still draws; the control fails signal a at Open trace, and signal b. Jake chose to amend the spec over adding a TEAMS control (decision on #2759). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016AYpa2x1a2xPUDxGueDEZP
|
The latest updates on your projects. Learn more about Vercel for GitHub. 3 Skipped Deployments
|
Codex Review SummaryThis 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. |
|
Second look / architecture pass: approve the approach, one optional nit. Approach. Correcting the spec to name the assertions the control actually fails is the leanest fix. Adding a TEAMS-specific control would test the page config twice, so I agree with the decision recorded in the PR body. No refactor opportunity here: it is a one-line spec change, with no code to consolidate. Nit (optional, take it only if you touch the file again). The new text names the assertion Also worth a glance. The mermaid at Generated by Claude Code |
There was a problem hiding this comment.
Problem this fixes
The verify table said GOAL_CONTROL=no-page-config must fail TEAMS, but TEAMS is driven by API reads on the page, not by injected page meta. That made the control easy to mis-implement and to “game” when S13 lands.
Verdict
Approve. This is the right-sized fix: one line, no new machinery, and it matches how controls are documented elsewhere (name the assertions that go red; partial leg-a failure is normal).
The amendment is slightly broader than a pure rename (it also names b under no-page-config, not only Open trace). That is coherent with PLAN S5’s boot-id-in-meta reload path for goal check b — without page meta, watch can still restart the Lab while the browser assertion for “the change shows” should fail. Worth one explicit tie to S5/BR-16 in follow-up prose so implementers do not treat b as no-watch-only.
/simplify — what I would not do
- Split verification into sub-legs a1/a2 in this PR — correct idea for readability, but scope creep for an amend.
- Add goal scaffolding here — S13 owns
goal.md/run.mts; SPEC should stay the short table.
Optional polish (non-blocking)
See inline. None change the direction.
Parallel review notes
- Code quality: no runtime code; table cell is dense but acceptable for a spec amend.
- Performance: n/a.
- Reuse: mirror the TEAMS-stays-green + assertion naming detail in
goals/shift-manager/it-runs-from-an-install/goal.mdwhen S13 ships (same pattern asit-opens-a-lab,kitchen-sink-talkcontrols).
Sent by Cursor Automation: Review For Best Approach


Requested by Jake · project thread
Before: FIX-1770's spec says the install check's
no-page-configcontrol must fail the sidebar's TEAMS list. It can't. TEAMS comes from the Lab's API, which still answers when the page config is gone.After: the spec names the assertions the control actually fails: Open trace opens the Lab's session (signal a) and the change shows (signal b). This is what
goals/shift-manager/it-runs-from-an-install/goal.mdalready says on #2759.How: one line in
specs/issues/FIX-1770/SPEC.md. Jake chose this over adding a TEAMS control, on the decision card in the FIX-1770 PR-C thread. Codex flagged the mismatch on #2759.No changeset needed: spec only.
🤖 Generated with Claude Code
https://claude.ai/code/session_01E4JkbyjudUYb1VEvw5wphU
Generated by Claude Code