Skip to content

docs(ledgers): T1 closes on corpay-console-smoke's first green night; G11 met (#325) - #583

Open
omercelikdev wants to merge 2 commits into
mainfrom
agent/325-t1-first-green-night
Open

omercelikdev wants to merge 2 commits into
mainfrom
agent/325-t1-first-green-night

Conversation

@omercelikdev

Copy link
Copy Markdown
Collaborator

What and why

#325's last open criterion (AC3 in #540) is closing T1, and T1's row names the proof that closes it: "the thread closes with the link to its first green night."

When #540 merged (2026-11-10 08:37Z), T1 could not close in that PR, because no night had run yet. That night has now run, and so have six more.

Nightly Date corpay-console-smoke Whole run
154 2026-11-10 success (job), the first nightly after #540 cancelled
155 2026-11-11 success failure
156 2026-11-12 success cancelled
157 2026-11-13 success failure
158 2026-11-14 success cancelled
159 2026-11-15 success cancelled
160 2026-11-16 success cancelled

The job ran for real, not as a skipped step. In run 154 the drive step took 5 min. CorPay was built on its published train (0 warnings), Keycloak 26.0.7 was pulled, and the Playwright log reads:

Running 4 tests using 1 worker
  ✓  1 e2e/corpay.spec.ts:49:3  › the API head's console holds its floor: no token is 401, a token without the ops role is 403
  ✓  2 e2e/corpay.spec.ts:66:3  › the EOD job on the internal head: run it, watch it in the history, pause and resume it
  ✓  3 e2e/corpay.spec.ts:127:3 › a payment file through the four-eyes gate: the maker uploads, the approver approves, each is on the record
  ✓  4 e2e/corpay.spec.ts:198:3 › the approved payment file executes, and today its row fails as #539 describes
  4 passed (3.3m)

Honesty about the whole runs. The nightly runs end cancelled or failure because of other jobs. The Messaging mutation job hits its 300-minute timeout (#562, PR #577 open). T1's proof is this job's night, and the ledger entry says exactly that, so nobody reads "the nightly was green".

The change (ledgers only; no package, script or workflow change)

  • docs/strategy/open-threads.md: the T1 row leaves the open table, and a Closed threads entry carries the job link, what the four specs proved, and the green streak. T62, T63 and T64, which the drive found, stay open in their own rows.
  • docs/strategy/ga-gate-status.md G11: now Met, with the link. This file holds gate status. docs/rfc/goldpath-ga-readiness.md is untouched, so the PR is not Tier C on that account.
  • docs/strategy/ai-sdlc-status.md: "T1 is still OPEN" becomes CLOSED, with the night.
  • docs/strategy/master-plan-2026-08.md item 11: a dated status note. The plan's original sentence is kept.
  • CHANGELOG.md [Unreleased] › Changed: one entry.

Reproduce

git fetch origin agent/325-t1-first-green-night && git checkout agent/325-t1-first-green-night
./scripts/docs-freshness.sh && ./scripts/ledger-check.sh && ./scripts/schema-honesty.sh && ./scripts/skills-parity.sh && ./scripts/template-pins.sh
# the evidence: the first nightly after #540 (merged 2026-11-10T08:37:34Z) and its corpay-console-smoke job
gh api repos/qorpe/goldpath/actions/runs/42789231201/jobs --jq '.jobs[] | select(.name=="corpay-console-smoke") | "\(.conclusion) \(.started_at)"'
gh api repos/qorpe/goldpath/actions/jobs/129738098092/logs | grep -E "✓|passed"
# red: put the T1 row back in the open table without removing the Closed entry, or point G11 at a job that failed —
# the ledgers then contradict each other; ledger-check (with an authenticated gh) is the gate that reconciles them

This is a ledger change, so there is no failing test to plant. The "red" here is the evidence itself: a night where the job failed or was skipped would leave T1 open. Every night since #540 is listed above with the job's own conclusion.

Self-check

  • Acceptance criterion (G11 (T1): the console smoke DRIVES CorPay's admin surface end to end #325 AC3): T1 closes with the link to its first green night. The link is above and in the Closed entry.
  • Ledgers: every ledger that said "T1 closes on the first green night" is updated. A grep -rn "first green night" over *.md now matches only the historical CHANGELOG entry for test(console): the console DRIVES CorPay through the consoles it serves, two principals via Keycloak (G11, T1, #325) #540 and the new entries. CHANGELOG [Unreleased] is updated. No new deferral, so no new T-number.
  • Honesty:
    • G11 says "Met" on the job's evidence and says nothing about the run as a whole.
    • The Closed entry says the runs end cancelled, and why.
    • The ai-sdlc row keeps its record of the 2026-09-05 withdrawn claim.
  • ga-readiness RFC untouched: status lives in ga-gate-status, per the owner decision of 2026-10-23.
  • Not applicable: XML docs, contracts, benches, wiring, security, concurrency and migrations. No code changes.
  • Confidentiality: no client, operator or brand names. "CorPay" is the repository's own sample.

Gates run

  • docs-freshness.sh rc 0, schema-honesty.sh rc 0, skills-parity.sh rc 0, template-pins.sh rc 0.
  • ledger-check.sh rc 0, but it printed "no authenticated gh — skipped honestly", so its GitHub reconciliation did not run here. That is the one gate I could not run. Please run it with an authenticated gh.
  • No build or tests: no code changed.

What I could not verify

Closes #325


Generated by Claude Code

… G11 met (#325)

The proof T1's row named is the link to the job's first green night: nightly run 154,
2026-11-10, the first nightly after #540 merged. It drove CorPay on its published train with
Keycloak 26.0.7, and 4 of 4 Playwright specs passed: the API head's floor, the EOD job on the
internal head, the four-eyes payment file by two principals, and the approved file pinned to
#539's failure shape. Green on every nightly since (runs 154-160). The nightly runs themselves end
cancelled for the Messaging mutation timeout (#562), not for this job.

open-threads: T1 to Closed threads. ga-gate-status: G11 met. ai-sdlc-status and the master
plan's item 11 say so. CHANGELOG [Unreleased].

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018AFdNZjRzMYQfNkJCg4Zjr
@omercelikdev omercelikdev added the agent-pr PR opened by the overnight agent label Nov 16, 2026
@github-actions github-actions Bot added the tier:A ADR-0014 Tier A: merges automatically once agent-verified and green label Nov 16, 2026
…ds no closed claim on open issues (#325)

ledger-check reads any line holding 'closed' as claiming every issue on it is closed. The G11 row
names #325 and #539, both open until this merges, and the master plan's note names #325.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018AFdNZjRzMYQfNkJCg4Zjr
@omercelikdev

Copy link
Copy Markdown
Collaborator Author

CI follow-up: gates failed on 14595fb. It was the one gate I could not run here: ledger-check.sh's GitHub half.

Cause: ledger-check reads any line containing "closed" as a claim that every issue number on that line is closed. Two lines tripped it:

Fix in 3514334: both lines are reworded without that word. G11 now says "T1 has left the open table of open-threads", and the master plan says "T1's proof ran on …". The meaning is unchanged. On 3514334, gates is success, and every check apart from integration (still running) is green.


Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

agent-pr PR opened by the overnight agent tier:A ADR-0014 Tier A: merges automatically once agent-verified and green

Projects

None yet

Development

Successfully merging this pull request may close these issues.

G11 (T1): the console smoke DRIVES CorPay's admin surface end to end

2 participants