test(testing): re-run the first-consumer pass for the visual-parity and feature-map skills on a host with a chromium build #7195

Activity

1. github-actions

   added

   needs-triageNot yet classified. Floor until a type and one priority tier are set.Not yet classified. Floor until a type and one priority tier are set.

   on Oct 11, 2026
2. github-actions commented on Oct 11, 2026

   

   github-actionsBot

   on Oct 11, 2026 – with GitHub Actions

   Contributor

   More actions

   Thanks for filing this. As I understand it: the visual-parity and feature-map skills from #7009 (/testing:map-features, /testing:refresh-feature-map, /testing:check-visual-parity) still need a first-consumer pass, since it was blocked on a WSL host where Playwright has no chromium build for ubuntu26.04. The ask is to re-run the pass on a host where playwright-cli resolves a chromium build, exercise all three skills end to end against a real app, and record the result (pass, or defects filed) on this issue.

   Scope and done criteria look clear and self-contained, with #7009 and #7137 already cross-referenced. I didn't find any existing open issue duplicating this one.

   _{Automated triage. A maintainer reviews it.}
3. github-actions

   added

   priority: mediumReal value, no hard deadline; normal backlog flow.Real value, no hard deadline; normal backlog flow.

   on Oct 11, 2026
4. kyle-sexton

   removed

   needs-triageNot yet classified. Floor until a type and one priority tier are set.Not yet classified. Floor until a type and one priority tier are set.

   on Oct 12, 2026
5. kyle-sexton

   added the

   Task

   issue type on Oct 12, 2026
6. kyle-sexton

   added a parent issue on Oct 16, 2026

   • app-control: new plugin plus testing and debugging changes #7581
7. kyle-sexton

   added

   needs-humanHuman-in-the-loop required; autonomous sessions must not resolve items carrying this.Human-in-the-loop required; autonomous sessions must not resolve items carrying this.

   on Oct 19, 2026
8. kyle-sexton commented on Oct 19, 2026

   

   kyle-sexton

   on Oct 19, 2026

   MemberAuthor

   More actions

   This was generated by AI during triage.

   Held: the first-consumer pass needs a host where playwright-cli resolves a chromium build; the WSL host has none for ubuntu26.04.
9. kyle-sexton commented on Oct 25, 2026

   

   kyle-sexton

   on Oct 25, 2026

   MemberAuthor

   More actions

   This was generated by AI during an unattended host run over /fleet:reach (melo-desk-001, native Windows lane).

   The blocker is gone on the desktop Windows lane. In Git Bash, playwright-cli 0.1.22 resolves at ~/.bun/bin/playwright-cli and opens a browser through the installed Chrome: playwright-cli open about:blank, then snapshot and close, all succeeded. No bundled chromium download was needed.

   The first-consumer pass did not run in this session. It needs a target app chosen. /testing:map-features writes a feature-map skill into that app's repository, and /testing:refresh-feature-map can open a PR there. Both are writes outside this repository that this host run was not scoped to make. Next step: name the app, then run the three skills on this lane through /fleet:reach.
10. kyle-sexton

    added sub-issues on Oct 26, 2026

    • fix(testing): say how check-visual-parity runs a suite and what a run with no verdict means #10105
    • fix(app-control): require a Ready command that fails when the instance is unfit #10145
11. kyle-sexton commented on Oct 26, 2026

    

    kyle-sexton

    on Oct 26, 2026

    MemberAuthor

    More actions

    First-consumer pass result (owner decision Q112, 2026-10-25): melodic-software/medley identity-server, native Windows on the desktop, launched with tests/Identity.E2E.Launcher (Testcontainers SQL Server on port 5099, mail discarded), driven with playwright-cli 0.1.22 and Chrome.

    Cost check before pushing to medley: no GitHub-hosted job runs for a draft pull request or a feature-branch push. The three ubuntu-24.04 jobs skip drafts (pr-require-checks.yml:1311,1353,1432). The review lanes skip drafts (pr-review.yml:72, pr-review-security.yml:66). Every other job and reusable call runs on melodic-ubuntu-24.04-x64, and every push trigger is limited to main. The draft's live run confirmed this: the branch push started 0 runs; the three ubuntu-24.04 jobs and both review lanes were skipped; ci-status refused the draft as designed.

    Results

    Skill	Outcome
    /app-control:map create	Control skill control-identity-server delivered in draft PR https://github.com/melodic-software/medley/pull/2222: a hub, 4 features, config at the safe floors, and a bash driver CLI. check against current main: 0 findings, after fixing two index-entry-count breaks it found (wrapped entry-point lines). CLI floors: local, environment and flag loosenings each exit 4. Drive proof of sign-in: Ready, then 5 steps as written, including one CLI sign-in attempt; one screenshot; cleanup matched by pid, start time and command-line hash; no orphan; all evidence non-empty
    /testing:check-visual-parity	Route 1, the repository suite: it failed twice on medley drift before comparing anything (SDK default 10.0.301 against a 10.0.401 pin; Playwright image 1.62.1 against @playwright/test ^1.64). With the runner's two override variables set, 5 of 5 states matched the Linux baselines. Route 2 (/app-control:drive capture): baseline for "Register, validation errors" plus one compare checkpoint, differing=0 tolerance=0 verdict=pass. Consent: gap, not reachable (below)
    /testing:map-features	Run first, before #10033 retired it. Its map went out in the PR's first commit; the second commit replaces it with the control skill

    The finding the research expected, "route 1 fails on win32 because the baselines are Linux-only", did not reproduce: medley's runner moves non-Linux hosts into its Linux container.

    Defects filed as slices

    • fix(testing): say how check-visual-parity runs a suite and what a run with no verdict means #10105: check-visual-parity route 1 does not say how to run the suite, and has no rule for a run that ends without a verdict.
    • fix(app-control): require a Ready command that fails when the instance is unfit #10145: nothing requires an app-control Ready command to fail when the instance is unfit.

    Not filed:

    • Drive's launch record and pid binding define only a Linux form. On native Windows this run recorded the owner SID, the creation time and a hash of the command line instead. That is the "Platform" bullet of fix(app-control): close the open threads pid-binding and eval follow-ups #9546.
    • Map's pointer-target rule put the pointer in medley's AGENTS.md, while medley's AGENTS.md:5 routes .claude/* paths to .claude/rules/. A placement conflict, with no harm shown yet.

    medley findings (in the PR body, not filed here)

    Over the HTTP launcher the session cookies (SameSite=None without Secure) are dropped, so no sign-in persists and Consent cannot be reached. The auth rate limit also covers /Error and /Consent. An unknown client lands on a 404. The suite runner's SDK and image defaults have drifted.