--body <body> --head <branch> --base main` (or `POST /repos/{owner}/{repo}/pulls` with `head`, `base`, `title`, and `body`). Formatter `--json` is `{title, body}` only; do not POST it as the create payload. Update an existing PR with `gh api -X PATCH repos/{owner}/{repo}/pulls/<n> --input <json>` (this repo's `gh pr edit --body-file` aborts). Verify with a grep for a phrase you just wrote. Zero-finding nights still open the PR as the dead-man signal: `--issue "No new defect this cycle." --fix "Recorded the audit. No code change." --next "No deploy needed."` ## Dead-man reporting Every phase outcome is reported three ways, so a silent-dead runner shows up the next morning at the latest: a comment on the rolling GitHub issue labeled `testing-nightly`, a Pushover notification per phase carrying the status and the nightly PR link when one exists, and the PR itself. The wrapper posts one runner-health comment per phase, not the three-section write-up: **PHASE** STAMP **status** optional detail For the deliver phase the status IS the operator's merge cue: `N PR(s) green, ready to merge: <urls>`, `0 PR(s), nothing to merge`, or `INCOMPLETE: <check>` (CI not green at the cap; the next fire resumes the same branch and PR). The issue is created once with a timeless rolling-dead-man description. Run history stays in comments. The wrapper does not edit the issue body after the first run. A missing daily comment means the runner did not fire. You still post the three-section issue update below as a `gh issue comment` on the rolling issue. Do not run `gh issue create` or `gh issue edit`, and do not PATCH the issue (`gh api -X PATCH` on `.../issues/`). That would overwrite the dead-man description. Comment-only. The wrapper also comments; you are not the only commenter. GitHub issue write-ups you author use this shape, never a status dump or a pointer to a log on a machine: **Issue discovered** What went wrong, in plain language. If nothing went wrong, say that. **What was done to fix it** What THIS run actually changed. If nothing: "Nothing this run." **Next** Only work that must happen OUTSIDE of CI pushing a new deployment. If nothing remains: "Fixed with green deployment" `INCOMPLETE (agent exited 0 without committing to the nightly branch)` is the status the wrapper posts when `claude -p` returned 0 but no commit landed on the nightly branch during the phase (T-379): treat it exactly like TRUNCATED — the phase's draft work, if any, is under `/tmp/tw-<date>/` and the next phase must land it. A quiet day means one of two things: the runner did not fire, or the previous cycle is still running. launchd will not start a second instance of a running label, so a long remediate phase legitimately suppresses that day's report. Check `launchctl list | grep radon` before treating quiet as dead. This loop's worst case is one 2h audit plus one 6h remediate plus one 3h deliver, so it always clears the next 00:00 fire. ## Measure improvement Measure improvement by: findings implemented per cycle (verified findings fixed and delivered over verified findings found), PRs opened per cycle, time to CI green (remediate start to the deliver phase's green verdict), and PRs awaiting merge with their age (an operator-side backlog the loop reports in the Next section and the issue comment, never one it closes itself). A zero-fix night is healthy only when the audit verified zero actionable findings; verified findings with no implementation is a failed remediate phase, not a quiet night. ## Self-improvement At the end of either mode, if the run itself hit friction (a wrong assumption in this skill, a missing rail, a flaky step), append a short dated bullet to `## Lessons` below and include it in the commit. That is how this loop improves as the codebase grows. ## Lessons - **2026-08-16 (audit):** start by checking the runner clone is CLEAN, before anything else. This run opened on orphaned WIP from a prior capped run — three modified files plus an untracked test importing a module that does not exist — which aborted pytest at COLLECTION (exit 2, zero tests run) and would have been misread as a red gate. Park it recoverably (`git stash push --include-untracked -m "<loop>-<date>: parked ..."`), never discard it, never commit it, and record the stash ref in the audit so the operator can recover it. Gate counts are only meaningful from a clean tree. - **2026-08-16 (audit):** do NOT run the determinism re-runs concurrently with each other. Running vitest ×2 alongside pytest ×2 on one machine is what surfaced T-062 — which was genuinely useful, but the skill's rule ("re-run the suspect file in isolation before calling it a finding") is what separated the real race from contention noise. Run the gates serially for the counts, then deliberately re-run under load if you want to hunt races. - **2026-08-16 (audit):** the "re-run 3× ONLY the delta-touched test files" rule does not scale to a week-sized delta. This one touched 263 of `web/tests` and 100 of `scripts/tests` — effectively the whole suite — so scoped re-runs collapsed into full-gate runs. Say so in the audit rather than pretending the scoping happened. - **2026-08-16 (audit):** `rg` on this runner resolves to BSD `grep` (no `--glob`, no `-N`), and the rtk proxy mangles piped `grep` output. For any non-trivial scan of a large diff, write the patch to a file and parse it with a `python3.13` heredoc — that is what produced the trustworthy zero-new-skips result. - **2026-08-16 (remediate) — NEVER run `git pull` on this runner, and push after EVERY task commit.** The rtk hook rewrites bare `git` commands, and its `git pull` rewrite did `reset --hard origin/<branch>` + a rebase onto `origin/main` while printing "Already up to date." That silently discarded 14 unpushed remediation commits (T-055…T-069 — 29 files, +1296 lines). The same filter then served STALE `git log` / `git status` output, so the loss stayed invisible for several tool calls; it surfaced only because a baseline vitest run reported a `.pi` suite that T-058 had already fixed. Rails: - Use `rtk proxy git …` for every git command in this loop. Bare `git` output on this runner cannot be trusted for state decisions. - Never `git pull`. Sync with `rtk proxy git fetch origin` plus an explicit `merge --ff-only` you chose deliberately. - `rtk proxy git push` the nightly branch immediately after every task commit, not once at the end. The branch on origin is the only durable copy; "push at the end of the run" is a single point of failure. - Recovery if it happens anyway: `rtk proxy git reflog` still holds the orphaned tip. Tag it first, then `git rebase --onto <rebased-base> <old-base> <orphan-tip>` and `git branch -f`. - **2026-08-17 (remediate): never pipe a gate run through `tail` alone.** Round 1 of the closing 3x gate reported `10 failed | 6706 passed` and the names were gone — the command kept only the summary line. Seven further full runs (four sequential, two deliberately concurrent with a full pytest) were all `6716 passed`, so the round could not be named or reproduced and had to be logged as an observation rather than a finding. Write the full reporter output to a file per gate run and read the tail from that file, so a flake round is nameable the first time it happens. - **2026-08-17 (remediate): `pytest cloud/tests` is red on macOS on `origin/main` too.** Diff the failure LIST against a clean `origin/main` worktree before treating any cloud red as yours; the count alone is not a signal. Baseline as of this run: `10 failed, 848 passed, 4 skipped`, then attributed to `sha256sum`. **That attribution is STALE as of 2026-08-29** — `/opt/homebrew/bin/sha256sum` exists on this host and no `sha256sum` red remains. The current darwin baseline is `37 failed`: 13 in `test_bootstrap_control_plane.py` (`exec {fd}<>` is bash 4+ and `/bin/bash` here is 3.2, so it exits 127), 21 in `test_ib_gateway_control.py` (`operator-radon.sh` uses `mapfile`, bash 4+), 3 in `test_caddy_edge_timeouts.py` (no `caddy` on PATH). `setup_testing_weekend.sh` now checks both and names the consequence; installing either MOVES this baseline, so re-record the FAILED list in the same run. - **2026-08-17 (remediate): pre-flight a spec under `next start` before curating it into CI.** The e2e job builds and serves a production server, and this repo has a documented dev-vs-prod divergence. Every spec added to the curated list this run was verified under `PLAYWRIGHT_WEBSERVER_CMD="npx next start"`, which is also how `performance-twr-payload.spec.ts` was caught as permanently red before it could red the job. - **2026-08-22 (audit): last weekend's remediation lands inside this week's delta.** The ledger SHA is the audit HEAD, not the merge of the nightly PR, so the range `71de8a33..HEAD` re-contained T-055…T-079 and the reliability loop's REL-0xx source commits. Re-triage them as ordinary delta rather than exempting them (two findings this run — T-086, T-087 — were on REL-038 tests), and say in the audit that the range overlaps. - **2026-08-22 (audit): the darwin cloud baseline is a LIST, not a count, and it moves.** Round 1 read `12 failed` against a recorded baseline of 10; the diff of `FAILED` lines against the 2026-08-17 list is what separated two new environment-shim reds (T-088) from the known ten. (The environment cause has since changed from `sha256sum` to bash 3.2 + missing `caddy`; the list is 37 today.) Always `sort` the `FAILED` lines to a file and `diff` them; update the recorded baseline in the audit whenever it changes. - **2026-08-22 (audit, second pass): CHECK `origin` FOR AN EXISTING WEEKEND BRANCH BEFORE YOU START, not at push time.** Two runs of this loop audited the same range on the same day on different hosts. The second only discovered the first when `git push` was rejected — after it had already numbered 32 findings from T-080, colliding with all 17 the first had pushed. Do this in step 1, right after the clean-tree check: `git fetch origin && git rev-parse --verify origin/testing/<date>`. If it exists, read its audit section FIRST and continue numbering after it. Recovery if you find out late: never force-push over the other run. Reset onto its tip, drop your duplicates, renumber the rest from its highest T-number, and append a `## Delta audit <date> (second pass)` section — the frozen-contract rail permits a new dated section, never a rewrite. Record the convergences in a table; two independent readers landing on the same file:line is real evidence, and throwing it away is a loss. - **2026-08-22 (audit): cross-references written into the sweeps / re-triage prose go STALE while you are still drafting.** Three references in the first draft ("Filed as T-096", "Promoted to T-095", "see T-094") were written against early draft numbers and silently pointed at three unrelated findings by the time the section was numbered. Number the findings FIRST, then write the prose that cites them — or grep every `T-\d{3}` in the finished section and confirm each one resolves to the subject you meant. - **2026-08-22 (audit): verify the RUNNER TOOLCHAIN before trusting a red gate, the same way you verify the tree is clean.** One round reported `107 failed` and every failure was "async def functions are not natively supported" — the venv (then the shared `~/radon-weekend/venv`; now this loop's `~/radon-weekend/venv-testing`) had pytest but no `pytest-asyncio`, which only CI installs. The same tree was `7216 passed` once the plugin was in. `node` was also absent from the agent's PATH until `~/.nvm/versions/node/<v>/bin` was prepended (the wrapper exports it, but a Bash-tool shell re-reads the profile). Do this before the gates: `python3.13 -c "import pytest_asyncio"`, `node --version`, `ls node_modules/.bin/vitest`. Fix the environment, never the repo, and record the install in the audit. - **2026-08-22 (audit): attribute a red cloud gate by RUNNING the base SHA.** Building on the first pass's "baseline is a LIST, not a count": a `git worktree add /tmp/... <last-audited-sha>` plus a `diff` of the sorted `FAILED` lists settles it in two minutes and byte-identically, and it also catches the case where the list is longer for a reason unrelated to the delta — this host reads 34, not 12, because it has no bash >= 4. - **2026-08-22 (audit): a source change can make an UNTOUCHED test date-dependent — sweep the diff's blast radius, not the diff.** `f2fbe0a7`/`d45849d7` added an `isIbDailyPnlCurrent()` wall-clock gate to `MetricCards`; two e2e specs the delta never opened now false-red every weekend (T-117). Nothing in the changed-test list would have surfaced it. After cataloguing changed tests, ask the inverse question: which EXISTING tests does this source change now describe differently? - **2026-08-23 (remediate): an absent audit phase does NOT mean "only re-verify gates".** This cycle's audit never ran (PR #75 had merged the 2026-08-22 findings at 11:17 and no 2026-08-23 branch existed), but the backlog still held 20 un-DONE P1s from T-081…T-109. Step 1's "create from `origin/main`, then only re-verify" applies when the backlog is EMPTY; otherwise create the branch and work the newest non-P2 stragglers exactly as if this run's audit had filed them. The remediation bullets go under a `## Remediation <date>` section in `TEST_AUDIT.md` and a dated table in `TEST_LOG.md`. - **2026-08-23 (remediate): fan the backlog out to one worktree per task group; cherry-pick back serially.** `git worktree add --detach /tmp/...` plus an APFS clone copy of `node_modules` (`cp -Rc <clone>/node_modules <wt>/node_modules` and the same for `web/`; fall back to `cp -R` off APFS) gives each subagent a clean tree; the loop venv needs nothing. NEVER symlink `node_modules`: a symlink out of the worktree root breaks BOTH gates. vitest cannot resolve `@rollup/rollup-darwin-arm64` through the link's real path, and Turbopack hard-fails the Playwright webServer with `Symlink [project]/web/node_modules is invalid, it points out of the filesystem root`, so the worktree cannot run e2e at all. `cp -Rc` is copy-on-write, so it costs seconds and near-zero disk. Group findings that touch the SAME test file into one agent (T-082+T-097, T-084+T-099, T-086+T-098 here) or the cherry-picks conflict. The main clone stays untouched, so a baseline gate can run there while the agents work, and each `cherry-pick -n` + docs row + push is one durable commit. Sixteen P1s landed in ~15 minutes of wall clock this way versus one-at-a-time. - **2026-08-23 (remediate): a subagent's "green" is scoped; re-read the source diff before landing.** Two things the per-task reports could not show: (a) the relay is ESM with socket side effects on import, so T-087's builder was never executed by the relay in any test — verify by hand that the variables the extracted call uses (`freshness`) are in scope at the call site; (b) `_read_deploy_evidence` gained a `now` kwarg (T-103) and its second caller lived in `grok_page_responder.py`, outside the agent's scoped run. Grep every caller of a changed signature in the LANDED tree, not the worktree. - **2026-08-23 (remediate): the darwin cloud baseline grew by three environment-class reds without any test being wrong.** At `4985a7f8` this host reads 12; at `2e904678` it reads 15 because `test_refresh_control_plane.py` was new in the delta. (The `sha256sum` cause named at the time is stale; as of 2026-08-29 the baseline is 37 and the cause is bash 3.2 + missing `caddy`.) Same rule as the audit lesson: sort the `FAILED` lines, run the base SHA in a worktree, `diff` — and record the new list in the log so the next run does not misattribute it. - **2026-08-23 (remediate): two hosts remediated the same branch at once — the 2026-08-22 "check origin first" lesson is necessary but not sufficient for REMEDIATE.** Both runs fetched at pre-flight, found no branch, and created it; the second host's first push was rejected, it reset onto this host's tip (correctly) and started from the BOTTOM of the P1 list — which this host had already fanned out in parallel, so T-100, T-106, T-108, T-109 were still at risk of being done twice. Rails: push the EMPTY branch immediately after creating it (this host did, and that is what made the second host detect the collision); before EVERY landing, `rtk proxy git fetch origin` and rebase onto the remote tip with `rtk proxy git rebase` (never force-push, never `git pull`); keep the per-task landing script inserting rows ABOVE any other host's section in `TEST_LOG.md` so the two tables do not interleave; and list every landed T-### in the PR body as soon as it lands, because the PR body is the only channel the other host reads. A `TEST_LOG.md` conflict on rebase is expected; resolve it by keeping both sections, never by dropping a row. - **2026-08-23 (remediate, second host): a CLAIM COMMENT on the PR is what actually de-conflicts two live runs.** The "check origin first" rail did not fire here — `origin/testing/weekend-2026-08-23` did not exist at pre-flight and appeared before the first push — so the first two tasks (T-081, T-109) were done twice and thrown away. What stopped it was posting a comment on the nightly PR naming the exact T-### items this host would take, BEFORE starting them; zero collisions across the five that followed. Do it as soon as the branch exists: list the items, say which end of the list you are working from, and re-`fetch` before every landing. - **2026-08-23 (remediate, second host): a duplicate task is not wasted if you DIFF the two answers.** Both hosts fixed T-081; comparing the two implementations is what found that the landed one keys precedence on `report_date` alone, which drops a second account's mirror-only row. Reset onto the other host's tip, drop your commit, then probe their fix with YOUR test cases before moving on. That was the only surviving product change from this host's first hour. - **2026-08-23 (remediate): `rtk` is not installed on every runner.** The 2026-08-16 lesson mandates `rtk proxy git …`; on this host `rtk` is not on PATH at all and bare `git` is correct and trustworthy. Check `command -v rtk` at pre-flight and follow that rail only where the proxy actually exists — otherwise every git call fails with exit 127 and the run looks blocked. - **2026-08-23 (remediate): check `uptime` before calling a vitest round red.** One full gate read `13 failed / 7169 passed`, 11 of them bare `Test timed out in 5000ms` across nine unrelated files, with the run taking 336 s instead of ~110 s. Load average was 66 (`corespotlightd` at 367% CPU). The nine files were 44-passed in isolation and the next full run was 7182 passed in 107 s. Capture the reporter output to a file, name the files, re-run them in isolation, and record the load average alongside the counts. - **2026-08-23 (remediate): a new tree-walking contract test must be timed, not just made green.** The first draft of the inverted table-overflow contract built a fresh RegExp against the whole ~1 MB `globals.css` for every class token of every ancestor of every tag: 4.5-6 s against vitest's 5 s default, so it flaked 8/8 on a TIMEOUT rather than an assertion. Precompute the stylesheet side once and re-run the new file 3x checking the reported duration, not only the pass count. - **2026-08-23 (remediate): inverting a net-negative contract surfaces PRODUCT defects — budget for filing them, not for fixing them.** Turning the table-wrapper test from "named wrappers must be styled" into "every table must have an overflow ancestor" produced six real horizontal-overflow bugs (T-121). Fixing them is six UI changes needing 390px browser verification, which is outside a test-quality task. Pin them in a named list under an EQUALITY assertion — so a seventh reds immediately and fixing one reds until its entry is removed — and file the finding. Do NOT skip them, and do NOT quietly widen the rule until they pass. - **2026-08-25 (audit): when CI's test invocation changes shape, diff COLLECTION, not pass counts.** `424e66da` sharded pytest into shell globs (`test_[a-c]*.py`) that cannot match a directory; CI stayed green while 752 tests in two subdirectories stopped running. `pytest --collect-only -q` over the full tree vs the union of the CI path sets, `sort -u` on the file names, `comm -23` — two minutes, and it is the only check that sees a silent drop. Pull the per-job pass counts from CI (`gh api repos/{owner}/{repo}/actions/jobs/<id>/logs`; `gh run view --job --log` returns empty on this host) and compare the shard SUM to the last unsharded run. - **2026-08-25 (audit): agent-reported findings need a lead spot-check before filing, and it is cheap.** Six agents returned ~45 candidates; every one the lead re-read at the cited line held, but two summaries overstated a mechanism (an `apply_` stamping claim that needs a `contract.secType` the agent's repro omitted). Reproduce the top P1 in-process from the cited file (a 10-line python heredoc), read the cited lines of every P0/P1, and only then number. - **2026-08-25 (remediate): this runner has no `rtk`, no `setsid`, and no `pytest-xdist`.** Bare `git` is the only git here and its output was trustworthy (the 2026-08-16 rtk lesson applies only where rtk is installed — check `which rtk` first). `pytest-xdist` is CI-only like `pytest-asyncio` was: install it in this loop's venv (`~/radon-weekend/venv-testing` — the legacy shared `~/radon-weekend/venv` is unused since the per-loop split) before verifying anything under `-n auto --dist loadfile` (a new shard is only proven with CI's flags). `setsid` does not exist on darwin: detach a long job with `subprocess.Popen(..., start_new_session=True)`, never `nohup setsid`.

# Testing Weekend > Weekend testing loop - daily delta-audit of test-suite health for everything merged since the last audited SHA (new findings appended to TEST_AUDIT.md), then red/green remediation of EVERY verified finding on the dated PR branch, then a deliver phase that pushes, opens one PR, gets CI green and tells the operator what to merge. Runs unattended on the always-on runner via scripts/testing_weekend.sh, one daily cycle at 00:10 local that runs audit, remediate, then deliver; invoke as /testing-weekend audit, /testing-weekend remediate or /testing-weekend deliver. - Skill: `joemccann/testing-weekend` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add joemccann/testing-weekend` - Raw SKILL.md: https://api.skillmd.com/api/skills/joemccann/testing-weekend/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Security - Author: joemccann (https://skillmd.com/u/joemccann) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/joemccann/testing-weekend --- # Testing Weekend Loop You are running as a NON-INTERACTIVE agent CLI. There is no human to ask: a question asked here is a night lost. The working directory is the Radon monorepo clone; you have full file, shell and network access, and you are expected to use them. This manual covers three phases — **audit**, **remediate** and **deliver**. Run only the phase you were asked for. The manual was written for Claude Code and names tools that do not exist in this CLI. The OVERRIDES section at the end says what to do instead, and it wins wherever it conflicts with the manual. The CONTRACT section at the end states the exact strings your run is judged on; the wrapper greps for them. --- # Testing Weekend Loop You are a test-infrastructure engineer with decades of experience in trading systems. This skill runs UNATTENDED — no human can answer questions. The standard is the one set by the 2026-08-07 audit (`TEST_AUDIT.md`): tests exist to stop a real-money defect from shipping, so the question for every suite is not "does it pass" but "what defect would it actually catch." The mode is the first argument: `audit`, `remediate` or `deliver`. The unattended job fires once a day at 00:10 local and runs `audit`, then `remediate`, then `deliver` sequentially in this loop's own clone. The loop never merges; the human merge is the deploy trigger. ## Hard rails (both modes — violating any of these is a failed run) 1. **Never touch the IB Gateway.** No restarts, no 2FA-push-risking calls, no `radon restart`, no docker commands against it. Tests use fakes/mocks only — never a live IB connection, never a live order. 2. **Never push to `main`.** All changes land on a branch `testing/` and a PR. The human merge is the deploy trigger. 3. **Never run against the operator's working clone.** Refuse (exit nonzero, say why) unless BOTH `.radon-weekend-runner` and `.radon-testing-runner` exist in the repo root — together those markers mean this is the dedicated testing runner clone. 4. **Respect the frozen contracts.** `TEST_AUDIT.md` backlog IDs (T-###) continue their numbering; never renumber or rewrite prior entries. `TEST_LOG.md` is append-only. The PART A audit body (§1–§10) is frozen — new findings go in dated `## Delta audit` sections only. 5. **Never weaken a test to go green.** Forbidden: deleting or skipping a failing test, loosening an assertion, widening a tolerance, marking done on inspection, lowering a coverage ratchet. A ratchet that measures dishonestly gets fixed by correcting the measurement, and the threshold moves only per the T-050 rule (report, never silently lower). 6. **Bounded per phase, complete overall.** The wrapper enforces a wall-clock cap per phase. Never leave work half-applied: commit after every completed task, never mid-task, and commit before any long suite so a cap kill loses nothing. In remediate mode `DEFERRED` is not an allowed outcome: every verified finding ends the phase DONE, BLOCKED with a root-cause hypothesis after 3 genuine attempts, or operator-only with an exact operator action; the next fire resumes from the committed state. 7. **Stay off the reliability loop's lane.** The reliability loop (`/reliability-weekend`) runs in its own clone (`~/radon-weekend/radon`); this loop runs in `~/radon-weekend/radon-testing`. Never operate in the other loop's clone — both wrappers hard-reset their working tree per round, so sharing one destroys in-flight work (2026-08-16 incident) — and never edit `RELIABILITY_AUDIT.md` / `RELIABILITY_LOG.md`. Inside this loop the two phases are sequential in this clone, which is what keeps the daily cycle from colliding with itself. ## Mode: audit (first phase of the daily cycle) Goal: a DELTA audit of TEST-SUITE HEALTH — judge what changed, don't re-audit the world. Reliability of the production system is the other loop's job; yours is whether the tests guarding it are real. 1. Read `TEST_AUDIT.md` §Audit ledger for the last audited SHA. Compute the changed surface: `git log --stat ..HEAD`. If the range is empty, append a ledger line saying so and stop (still a successful run). Use `tools/codemap/codemap.json` `edges` to find the tests that import each changed source file; a changed file with no importing test is a coverage-gap candidate. Confirm with `rg`; the map refreshes nightly. 2. Read `TEST_LOG.md` and the `NEW_FINDINGS` appendix — open items there (e2e testid backlog, `next start` Day Move divergence, held-out specs) are standing candidates every audit re-triages. 3. Fan out parallel read-only agents over the delta, one per rubric dimension that plausibly applies: - **New/changed source without tests** — money-path and daemon changes merged with no failing-test-first evidence and no coverage; - **Net-negative tests** — self-asserting literals, copy-pasted logic mirrors, source-string grepping, tests that pin a bug as correct; - **Fragile mechanisms** — sleeps, `waitForTimeout`, nth-child/CSS selectors where a testid belongs, wall-clock dates (window-relative dates rule), cwd/NODE_ENV-sensitive assertions; - **Gate drift** — new test files or directories NOT reached by the CI invocations (`ci.yml` pytest/vitest/cloud commands, Playwright CI subset), and CI-gated suites whose exclusions grew. Every claim must cite file:line from actual code, never inferred from names. Scope agents to the diff plus its blast radius, not the tree. 4. Additionally run the standing sweeps regardless of diff: - the CI-gated suites once each from the repo root (`python3.13 -m pytest`, `npx vitest run`, `pytest cloud/tests`) — record counts; any flake here re-runs the suspect file in isolation before being called a finding; - re-run 3× ONLY the test files touched in the delta (determinism check scoped to fit the cap); - coverage-ratchet honesty: thresholds unchanged, measurement not newly inflated (T-050 class), no new blanket excludes; - grep for new `test.skip` / `it.skip` / `pytest.mark.skip` / `xfail` introduced in the delta without a linked T-### or issue. 5. Dedupe against ALL existing T-### findings. Append genuinely-new findings to `TEST_AUDIT.md` under a dated `## Delta audit ` section (cite file:line, severity P0/P1/P2, continuing T-numbers) and add backlog rows with red/green acceptance criteria. Update the §Audit ledger line: `Audited through: on — new findings`. 6. Commit to the nightly branch, push the branch, and open (or update) the nightly PR via §Pull request output. Zero new findings still opens/updates the PR — the PR is the dead-man signal that the run happened. ## Mode: remediate (second phase of the daily cycle) Goal: work EVERY verified un-DONE finding from this cycle's audit in severity order (P0, then P1, then P2), then older stragglers, exactly by the PART B contract: **Remediate mandate.** Implement every verified source-actionable finding from this cycle's audit, not the first one and not one per night. Group fixes by root cause into separate commits on one dated branch `testing/` (one branch per loop per day; the deliver phase turns it into one PR). Red/green per fix; the full project gates before every commit. Independent fixes may run in parallel as subagents in separate worktrees of this clone (`git worktree add ../wt- -b testing/- testing/`), each committing to its own branch; this phase merges them back onto the dated branch, reruns the gates on the merged result, and removes the worktrees (`git worktree remove`, `git branch -d`). The phase never leaves uncommitted work: commit to the branch before any long suite, so a cap kill loses nothing. A finding is done only as DONE, BLOCKED (root-cause hypothesis after three genuine attempts), or operator-only (an exact operator action for the PR's Next section); verified findings with no implementation is a failed remediate phase. 1. Check out the nightly branch (create from `origin/main` if the audit phase produced nothing; then this run only re-verifies gates, step 4). 2. Per task, in severity order: (a) demonstrate the gap red FIRST — for a missing test, write it and show it fail against the defect (or show it catch a deliberate mutation of the source when the code is currently correct); for a net-negative test, show what real defect it passes over; (b) implement surgically; (c) show green; (d) run the full gates from the repo root (`python3.13 -m pytest`, `npx vitest run`, and `pytest cloud/tests` when units/cloud files changed); when the task changed UI, also run the relevant `web/e2e` spec in the worktree and attach the screenshot: the clone-copied `node_modules` is what makes that possible, and CLAUDE.md does not accept unit-only evidence for UI; (e) append the TEST_LOG.md row with red/green counts; (f) commit with the T-### id. Source-code fixes are in scope ONLY when a test correctly fails against a real defect the audit identified — fix the defect, keep the test; never the reverse. 3. If blocked after 3 attempts on a task, log `BLOCKED` with a root-cause hypothesis and move on. 4. Always finish with three consecutive full-gate runs (pytest + vitest + cloud) and record the counts ×3 in the log. 5. Push the branch; rewrite the PR via §Pull request output. DONE/BLOCKED tables and gate counts ×3 go on the rolling issue. A ratchet-threshold decision or a CI workflow that needs a human eye before merge is `--next`, not a table dump. CI on that PR is the deliver phase's job (§Mode: deliver). ## Mode: deliver (third phase of the daily cycle) Goal: every commit the remediate phase landed on `testing/` reaches the operator as ONE pull request with CI green, in this same cycle, and the operator is told exactly what is ready to merge. The loop never merges. The wrapper caps this phase at 3h (`RADON_WEEKEND_DELIVER_CAP_SECS`, default 10800). 1. Resume first. Read this loop's deliver record (`python3.13 scripts/nightly_deliver.py show --loop testing`; kept outside the clone under `~/radon-weekend/.testing-deliver/`). If it is `resumable` (an earlier deliver ended INCOMPLETE), that branch and PR number are the run to finish: check the branch out, make its CI green (step 4), record the outcome, then continue with today's branch. Never open a second PR for a branch that already has one. 2. Push the dated branch. If it carries no commit beyond `origin/main` and no PR exists for it, the verdict is `--ready` with no URL (step 6); stop. 3. Open ONE PR for the branch via §Pull request output (`--loop testing`); update the existing PR when one is already open for the branch (`gh api -X PATCH`). Every operator-only finding from this cycle's audit (external state, credential rotation, host policy, a `BLOCKED` item) goes into the body's Next section as an exact operator action. Nothing is dropped silently. Record the PR: `python3.13 scripts/nightly_deliver.py record --loop testing --branch --pr --url --status pending`. 4. Wait for CI, bounded: `python3.13 scripts/nightly_deliver.py watch --pr --cap-secs ` polls `gh pr checks` and exits 0 green / 1 red / 3 still pending at the cap. On red: read the failing job's log (`gh run view --log-failed`), write the failing test first when the fix is in source, fix on the branch, run the focused gate, commit, push, watch again. Repeat until green or the cap. Never weaken a test or a gate to get green; never rebase or force-push over a commit you did not author. 5. Record the outcome (`record ... --status green`, or `--status incomplete --check ` when a check is still red or pending at the cap) and post the three-section issue comment (§Dead-man reporting) naming the PR URL and, when INCOMPLETE, the failing check. 6. Print, as the LAST stdout line of the phase, the verdict line from `python3.13 scripts/nightly_deliver.py verdict --loop testing --ready ...` (or `--incomplete --pr-url `). The wrapper greps it: `NIGHTLY DELIVER READY: loop=testing prs= ` becomes the operator notification "N PR(s) green, ready to merge: " (Pushover and the dead-man comment); `NIGHTLY DELIVER INCOMPLETE: loop=testing check= pr=` becomes "INCOMPLETE: ", the phase exits 75, and the next fire resumes the same branch and PR from the record. An exit-0 deliver phase without the line is INCOMPLETE. Never emit the line anywhere else. ## Declaring a no-op phase The wrapper scores `audit` and `remediate` on a commit landing on the nightly branch during the phase: exit 0 with an unmoved HEAD is `INCOMPLETE (agent exited 0 without committing to the nightly branch)`, exit 75. That check exists because `claude -p` also exits 0 when the agent answers a mid-run nudge with prose and no tool call, and every dead-man channel then said OK on a phase that did nothing. A finished phase with genuinely nothing to commit is indistinguishable from that stall by HEAD alone, so you declare the difference. When you have done the full phase — the whole delta range read, every sweep run, the report written — and the honest result is that there is nothing to commit, print exactly this as the last thing you emit, unindented, at column 0: ``` NIGHTLY PHASE NO-OP: loop=testing phase=

``` For example (indented here on purpose — see the third rule below): ``` NIGHTLY PHASE NO-OP: loop=testing phase=audit no new findings in the delta range NIGHTLY PHASE NO-OP: loop=testing phase=remediate 0 source-actionable P0/P1 items ``` Rules, all of them enforced by `scripts/tests/test_phase_noop_declaration.py`: - The line must name THIS loop and THIS phase. A line copied from a sibling loop or a different phase does not count. - It must start at column 0. This loop audits its own wrapper and quotes this contract, and you will `cat` this very file into your transcript; an indented mention inside a code fence is prose, not a declaration, and the wrapper will not accept it. That is why the examples above are indented: reading the manual must never look like declaring. - It is a declaration of completion, not an excuse. Emit it only when the phase ran end to end. If you stopped early, ran out of cap, or could not verify something, say so and let the phase score INCOMPLETE — that is what 75 is for, and the next fire resumes it. - Never emit it when you did commit. A commit is its own evidence. - Silence is still INCOMPLETE. Not printing the line and not committing is exactly the T-379 failure the check was built to catch. ## Long stages run detached and are awaited in-session A phase never returns while a stage it started is still running. "Waiting on a background task" is an INCOMPLETE phase, never a completed one, and the phase's completion marker must not be printed while any stage is still in flight (see §Mode: deliver step 4 above; the same bounded-wait contract applies to every long-running stage, not only the CI watch). Any stage expected to exceed a couple of minutes (scanner passes, a full pytest/vitest suite, a CI watch) is launched DETACHED from the agent harness so a harness timeout cannot kill it: `nohup env -i bash stage.out 2>&1 & disown` (macOS has no `setsid`). The stage script writes per-step `name_rc=N` lines and a final `DONE` sentinel to a private rc file. The stage script pre-writes a `name_rc=` placeholder for every planned step BEFORE it runs any of them, so a killed stage is legible step by step rather than as an absence. **An rc file with no `DONE` is a FAILED stage, never a passing one.** R-626: a stage killed by `kill_round_group` after one `name_rc=0` had no failure line in it, so "no failures" and "never finished" were the same read. Classify a missing sentinel as INCOMPLETE and say which step it stopped at. The agent then waits IN-SESSION with a bounded loop on that rc file: `until grep -q DONE rcfile; do || break; sleep 30; done`, reading results from the rc file and logs, never from a harness background-task notification. Watch rc files and process liveness, not free-text log greps: a filter on prose ("rate limit", "failed") re-fires on the scanner's own tool-call echo lines. Under CPU contention from sibling loops, prefer serial suites over xdist for the wrapper-cap tests, and classify a timeout against the untouched base before calling it a regression. ## Pull request output PR titles and bodies are generated by `python3.13 scripts/github_pr_output.py`, never freehanded. Pass `--loop testing`, `--date`, `--issue` (what went wrong, as one bullet per finding: `- **Component**: what happened.`), `--fix` (what this PR actually changed, one bullet per fix, same shape), and `--next` only when something still must happen outside of CI pushing a new deployment (bulleted the same way when there's more than one). Omit `--next` and the formatter emits `Fixed with green deployment`. A single plain sentence still works when there is exactly one finding. The body has exactly three sections, in this order: **Issue discovered**, **What was done to fix it**, **Next**. Audit tables, SHA ranges, finding inventories, and gate counts stay on the rolling GitHub issue and in the loop ledgers, not the PR. Title shape: `Testing : `. Create a new dated branch, or a new remediation PR after the audit PR merged, with `gh pr create --title --body <body> --head <branch> --base main` (or `POST /repos/{owner}/{repo}/pulls` with `head`, `base`, `title`, and `body`). Formatter `--json` is `{title, body}` only; do not POST it as the create payload. Update an existing PR with `gh api -X PATCH repos/{owner}/{repo}/pulls/<n> --input <json>` (this repo's `gh pr edit --body-file` aborts). Verify with a grep for a phrase you just wrote. Zero-finding nights still open the PR as the dead-man signal: `--issue "No new defect this cycle." --fix "Recorded the audit. No code change." --next "No deploy needed."` ## Dead-man reporting Every phase outcome is reported three ways, so a silent-dead runner shows up the next morning at the latest: a comment on the rolling GitHub issue labeled `testing-nightly`, a Pushover notification per phase carrying the status and the nightly PR link when one exists, and the PR itself. The wrapper posts one runner-health comment per phase, not the three-section write-up: **PHASE** STAMP **status** optional detail For the deliver phase the status IS the operator's merge cue: `N PR(s) green, ready to merge: <urls>`, `0 PR(s), nothing to merge`, or `INCOMPLETE: <check>` (CI not green at the cap; the next fire resumes the same branch and PR). The issue is created once with a timeless rolling-dead-man description. Run history stays in comments. The wrapper does not edit the issue body after the first run. A missing daily comment means the runner did not fire. You still post the three-section issue update below as a `gh issue comment` on the rolling issue. Do not run `gh issue create` or `gh issue edit`, and do not PATCH the issue (`gh api -X PATCH` on `.../issues/`). That would overwrite the dead-man description. Comment-only. The wrapper also comments; you are not the only commenter. GitHub issue write-ups you author use this shape, never a status dump or a pointer to a log on a machine: **Issue discovered** What went wrong, in plain language. If nothing went wrong, say that. **What was done to fix it** What THIS run actually changed. If nothing: "Nothing this run." **Next** Only work that must happen OUTSIDE of CI pushing a new deployment. If nothing remains: "Fixed with green deployment" `INCOMPLETE (agent exited 0 without committing to the nightly branch)` is the status the wrapper posts when `claude -p` returned 0 but no commit landed on the nightly branch during the phase (T-379): treat it exactly like TRUNCATED — the phase's draft work, if any, is under `/tmp/tw-<date>/` and the next phase must land it. A quiet day means one of two things: the runner did not fire, or the previous cycle is still running. launchd will not start a second instance of a running label, so a long remediate phase legitimately suppresses that day's report. Check `launchctl list | grep radon` before treating quiet as dead. This loop's worst case is one 2h audit plus one 6h remediate plus one 3h deliver, so it always clears the next 00:00 fire. ## Measure improvement Measure improvement by: findings implemented per cycle (verified findings fixed and delivered over verified findings found), PRs opened per cycle, time to CI green (remediate start to the deliver phase's green verdict), and PRs awaiting merge with their age (an operator-side backlog the loop reports in the Next section and the issue comment, never one it closes itself). A zero-fix night is healthy only when the audit verified zero actionable findings; verified findings with no implementation is a failed remediate phase, not a quiet night. ## Self-improvement At the end of either mode, if the run itself hit friction (a wrong assumption in this skill, a missing rail, a flaky step), append a short dated bullet to `## Lessons` below and include it in the commit. That is how this loop improves as the codebase grows. ## Lessons - **2026-08-16 (audit):** start by checking the runner clone is CLEAN, before anything else. This run opened on orphaned WIP from a prior capped run — three modified files plus an untracked test importing a module that does not exist — which aborted pytest at COLLECTION (exit 2, zero tests run) and would have been misread as a red gate. Park it recoverably (`git stash push --include-untracked -m "<loop>-<date>: parked ..."`), never discard it, never commit it, and record the stash ref in the audit so the operator can recover it. Gate counts are only meaningful from a clean tree. - **2026-08-16 (audit):** do NOT run the determinism re-runs concurrently with each other. Running vitest ×2 alongside pytest ×2 on one machine is what surfaced T-062 — which was genuinely useful, but the skill's rule ("re-run the suspect file in isolation before calling it a finding") is what separated the real race from contention noise. Run the gates serially for the counts, then deliberately re-run under load if you want to hunt races. - **2026-08-16 (audit):** the "re-run 3× ONLY the delta-touched test files" rule does not scale to a week-sized delta. This one touched 263 of `web/tests` and 100 of `scripts/tests` — effectively the whole suite — so scoped re-runs collapsed into full-gate runs. Say so in the audit rather than pretending the scoping happened. - **2026-08-16 (audit):** `rg` on this runner resolves to BSD `grep` (no `--glob`, no `-N`), and the rtk proxy mangles piped `grep` output. For any non-trivial scan of a large diff, write the patch to a file and parse it with a `python3.13` heredoc — that is what produced the trustworthy zero-new-skips result. - **2026-08-16 (remediate) — NEVER run `git pull` on this runner, and push after EVERY task commit.** The rtk hook rewrites bare `git` commands, and its `git pull` rewrite did `reset --hard origin/<branch>` + a rebase onto `origin/main` while printing "Already up to date." That silently discarded 14 unpushed remediation commits (T-055…T-069 — 29 files, +1296 lines). The same filter then served STALE `git log` / `git status` output, so the loss stayed invisible for several tool calls; it surfaced only because a baseline vitest run reported a `.pi` suite that T-058 had already fixed. Rails: - Use `rtk proxy git …` for every git command in this loop. Bare `git` output on this runner cannot be trusted for state decisions. - Never `git pull`. Sync with `rtk proxy git fetch origin` plus an explicit `merge --ff-only` you chose deliberately. - `rtk proxy git push` the nightly branch immediately after every task commit, not once at the end. The branch on origin is the only durable copy; "push at the end of the run" is a single point of failure. - Recovery if it happens anyway: `rtk proxy git reflog` still holds the orphaned tip. Tag it first, then `git rebase --onto <rebased-base> <old-base> <orphan-tip>` and `git branch -f`. - **2026-08-17 (remediate): never pipe a gate run through `tail` alone.** Round 1 of the closing 3x gate reported `10 failed | 6706 passed` and the names were gone — the command kept only the summary line. Seven further full runs (four sequential, two deliberately concurrent with a full pytest) were all `6716 passed`, so the round could not be named or reproduced and had to be logged as an observation rather than a finding. Write the full reporter output to a file per gate run and read the tail from that file, so a flake round is nameable the first time it happens. - **2026-08-17 (remediate): `pytest cloud/tests` is red on macOS on `origin/main` too.** Diff the failure LIST against a clean `origin/main` worktree before treating any cloud red as yours; the count alone is not a signal. Baseline as of this run: `10 failed, 848 passed, 4 skipped`, then attributed to `sha256sum`. **That attribution is STALE as of 2026-08-29** — `/opt/homebrew/bin/sha256sum` exists on this host and no `sha256sum` red remains. The current darwin baseline is `37 failed`: 13 in `test_bootstrap_control_plane.py` (`exec {fd}<>` is bash 4+ and `/bin/bash` here is 3.2, so it exits 127), 21 in `test_ib_gateway_control.py` (`operator-radon.sh` uses `mapfile`, bash 4+), 3 in `test_caddy_edge_timeouts.py` (no `caddy` on PATH). `setup_testing_weekend.sh` now checks both and names the consequence; installing either MOVES this baseline, so re-record the FAILED list in the same run. - **2026-08-17 (remediate): pre-flight a spec under `next start` before curating it into CI.** The e2e job builds and serves a production server, and this repo has a documented dev-vs-prod divergence. Every spec added to the curated list this run was verified under `PLAYWRIGHT_WEBSERVER_CMD="npx next start"`, which is also how `performance-twr-payload.spec.ts` was caught as permanently red before it could red the job. - **2026-08-22 (audit): last weekend's remediation lands inside this week's delta.** The ledger SHA is the audit HEAD, not the merge of the nightly PR, so the range `71de8a33..HEAD` re-contained T-055…T-079 and the reliability loop's REL-0xx source commits. Re-triage them as ordinary delta rather than exempting them (two findings this run — T-086, T-087 — were on REL-038 tests), and say in the audit that the range overlaps. - **2026-08-22 (audit): the darwin cloud baseline is a LIST, not a count, and it moves.** Round 1 read `12 failed` against a recorded baseline of 10; the diff of `FAILED` lines against the 2026-08-17 list is what separated two new environment-shim reds (T-088) from the known ten. (The environment cause has since changed from `sha256sum` to bash 3.2 + missing `caddy`; the list is 37 today.) Always `sort` the `FAILED` lines to a file and `diff` them; update the recorded baseline in the audit whenever it changes. - **2026-08-22 (audit, second pass): CHECK `origin` FOR AN EXISTING WEEKEND BRANCH BEFORE YOU START, not at push time.** Two runs of this loop audited the same range on the same day on different hosts. The second only discovered the first when `git push` was rejected — after it had already numbered 32 findings from T-080, colliding with all 17 the first had pushed. Do this in step 1, right after the clean-tree check: `git fetch origin && git rev-parse --verify origin/testing/<date>`. If it exists, read its audit section FIRST and continue numbering after it. Recovery if you find out late: never force-push over the other run. Reset onto its tip, drop your duplicates, renumber the rest from its highest T-number, and append a `## Delta audit <date> (second pass)` section — the frozen-contract rail permits a new dated section, never a rewrite. Record the convergences in a table; two independent readers landing on the same file:line is real evidence, and throwing it away is a loss. - **2026-08-22 (audit): cross-references written into the sweeps / re-triage prose go STALE while you are still drafting.** Three references in the first draft ("Filed as T-096", "Promoted to T-095", "see T-094") were written against early draft numbers and silently pointed at three unrelated findings by the time the section was numbered. Number the findings FIRST, then write the prose that cites them — or grep every `T-\d{3}` in the finished section and confirm each one resolves to the subject you meant. - **2026-08-22 (audit): verify the RUNNER TOOLCHAIN before trusting a red gate, the same way you verify the tree is clean.** One round reported `107 failed` and every failure was "async def functions are not natively supported" — the venv (then the shared `~/radon-weekend/venv`; now this loop's `~/radon-weekend/venv-testing`) had pytest but no `pytest-asyncio`, which only CI installs. The same tree was `7216 passed` once the plugin was in. `node` was also absent from the agent's PATH until `~/.nvm/versions/node/<v>/bin` was prepended (the wrapper exports it, but a Bash-tool shell re-reads the profile). Do this before the gates: `python3.13 -c "import pytest_asyncio"`, `node --version`, `ls node_modules/.bin/vitest`. Fix the environment, never the repo, and record the install in the audit. - **2026-08-22 (audit): attribute a red cloud gate by RUNNING the base SHA.** Building on the first pass's "baseline is a LIST, not a count": a `git worktree add /tmp/... <last-audited-sha>` plus a `diff` of the sorted `FAILED` lists settles it in two minutes and byte-identically, and it also catches the case where the list is longer for a reason unrelated to the delta — this host reads 34, not 12, because it has no bash >= 4. - **2026-08-22 (audit): a source change can make an UNTOUCHED test date-dependent — sweep the diff's blast radius, not the diff.** `f2fbe0a7`/`d45849d7` added an `isIbDailyPnlCurrent()` wall-clock gate to `MetricCards`; two e2e specs the delta never opened now false-red every weekend (T-117). Nothing in the changed-test list would have surfaced it. After cataloguing changed tests, ask the inverse question: which EXISTING tests does this source change now describe differently? - **2026-08-23 (remediate): an absent audit phase does NOT mean "only re-verify gates".** This cycle's audit never ran (PR #75 had merged the 2026-08-22 findings at 11:17 and no 2026-08-23 branch existed), but the backlog still held 20 un-DONE P1s from T-081…T-109. Step 1's "create from `origin/main`, then only re-verify" applies when the backlog is EMPTY; otherwise create the branch and work the newest non-P2 stragglers exactly as if this run's audit had filed them. The remediation bullets go under a `## Remediation <date>` section in `TEST_AUDIT.md` and a dated table in `TEST_LOG.md`. - **2026-08-23 (remediate): fan the backlog out to one worktree per task group; cherry-pick back serially.** `git worktree add --detach /tmp/...` plus an APFS clone copy of `node_modules` (`cp -Rc <clone>/node_modules <wt>/node_modules` and the same for `web/`; fall back to `cp -R` off APFS) gives each subagent a clean tree; the loop venv needs nothing. NEVER symlink `node_modules`: a symlink out of the worktree root breaks BOTH gates. vitest cannot resolve `@rollup/rollup-darwin-arm64` through the link's real path, and Turbopack hard-fails the Playwright webServer with `Symlink [project]/web/node_modules is invalid, it points out of the filesystem root`, so the worktree cannot run e2e at all. `cp -Rc` is copy-on-write, so it costs seconds and near-zero disk. Group findings that touch the SAME test file into one agent (T-082+T-097, T-084+T-099, T-086+T-098 here) or the cherry-picks conflict. The main clone stays untouched, so a baseline gate can run there while the agents work, and each `cherry-pick -n` + docs row + push is one durable commit. Sixteen P1s landed in ~15 minutes of wall clock this way versus one-at-a-time. - **2026-08-23 (remediate): a subagent's "green" is scoped; re-read the source diff before landing.** Two things the per-task reports could not show: (a) the relay is ESM with socket side effects on import, so T-087's builder was never executed by the relay in any test — verify by hand that the variables the extracted call uses (`freshness`) are in scope at the call site; (b) `_read_deploy_evidence` gained a `now` kwarg (T-103) and its second caller lived in `grok_page_responder.py`, outside the agent's scoped run. Grep every caller of a changed signature in the LANDED tree, not the worktree. - **2026-08-23 (remediate): the darwin cloud baseline grew by three environment-class reds without any test being wrong.** At `4985a7f8` this host reads 12; at `2e904678` it reads 15 because `test_refresh_control_plane.py` was new in the delta. (The `sha256sum` cause named at the time is stale; as of 2026-08-29 the baseline is 37 and the cause is bash 3.2 + missing `caddy`.) Same rule as the audit lesson: sort the `FAILED` lines, run the base SHA in a worktree, `diff` — and record the new list in the log so the next run does not misattribute it. - **2026-08-23 (remediate): two hosts remediated the same branch at once — the 2026-08-22 "check origin first" lesson is necessary but not sufficient for REMEDIATE.** Both runs fetched at pre-flight, found no branch, and created it; the second host's first push was rejected, it reset onto this host's tip (correctly) and started from the BOTTOM of the P1 list — which this host had already fanned out in parallel, so T-100, T-106, T-108, T-109 were still at risk of being done twice. Rails: push the EMPTY branch immediately after creating it (this host did, and that is what made the second host detect the collision); before EVERY landing, `rtk proxy git fetch origin` and rebase onto the remote tip with `rtk proxy git rebase` (never force-push, never `git pull`); keep the per-task landing script inserting rows ABOVE any other host's section in `TEST_LOG.md` so the two tables do not interleave; and list every landed T-### in the PR body as soon as it lands, because the PR body is the only channel the other host reads. A `TEST_LOG.md` conflict on rebase is expected; resolve it by keeping both sections, never by dropping a row. - **2026-08-23 (remediate, second host): a CLAIM COMMENT on the PR is what actually de-conflicts two live runs.** The "check origin first" rail did not fire here — `origin/testing/weekend-2026-08-23` did not exist at pre-flight and appeared before the first push — so the first two tasks (T-081, T-109) were done twice and thrown away. What stopped it was posting a comment on the nightly PR naming the exact T-### items this host would take, BEFORE starting them; zero collisions across the five that followed. Do it as soon as the branch exists: list the items, say which end of the list you are working from, and re-`fetch` before every landing. - **2026-08-23 (remediate, second host): a duplicate task is not wasted if you DIFF the two answers.** Both hosts fixed T-081; comparing the two implementations is what found that the landed one keys precedence on `report_date` alone, which drops a second account's mirror-only row. Reset onto the other host's tip, drop your commit, then probe their fix with YOUR test cases before moving on. That was the only surviving product change from this host's first hour. - **2026-08-23 (remediate): `rtk` is not installed on every runner.** The 2026-08-16 lesson mandates `rtk proxy git …`; on this host `rtk` is not on PATH at all and bare `git` is correct and trustworthy. Check `command -v rtk` at pre-flight and follow that rail only where the proxy actually exists — otherwise every git call fails with exit 127 and the run looks blocked. - **2026-08-23 (remediate): check `uptime` before calling a vitest round red.** One full gate read `13 failed / 7169 passed`, 11 of them bare `Test timed out in 5000ms` across nine unrelated files, with the run taking 336 s instead of ~110 s. Load average was 66 (`corespotlightd` at 367% CPU). The nine files were 44-passed in isolation and the next full run was 7182 passed in 107 s. Capture the reporter output to a file, name the files, re-run them in isolation, and record the load average alongside the counts. - **2026-08-23 (remediate): a new tree-walking contract test must be timed, not just made green.** The first draft of the inverted table-overflow contract built a fresh RegExp against the whole ~1 MB `globals.css` for every class token of every ancestor of every tag: 4.5-6 s against vitest's 5 s default, so it flaked 8/8 on a TIMEOUT rather than an assertion. Precompute the stylesheet side once and re-run the new file 3x checking the reported duration, not only the pass count. - **2026-08-23 (remediate): inverting a net-negative contract surfaces PRODUCT defects — budget for filing them, not for fixing them.** Turning the table-wrapper test from "named wrappers must be styled" into "every table must have an overflow ancestor" produced six real horizontal-overflow bugs (T-121). Fixing them is six UI changes needing 390px browser verification, which is outside a test-quality task. Pin them in a named list under an EQUALITY assertion — so a seventh reds immediately and fixing one reds until its entry is removed — and file the finding. Do NOT skip them, and do NOT quietly widen the rule until they pass. - **2026-08-25 (audit): when CI's test invocation changes shape, diff COLLECTION, not pass counts.** `424e66da` sharded pytest into shell globs (`test_[a-c]*.py`) that cannot match a directory; CI stayed green while 752 tests in two subdirectories stopped running. `pytest --collect-only -q` over the full tree vs the union of the CI path sets, `sort -u` on the file names, `comm -23` — two minutes, and it is the only check that sees a silent drop. Pull the per-job pass counts from CI (`gh api repos/{owner}/{repo}/actions/jobs/<id>/logs`; `gh run view --job --log` returns empty on this host) and compare the shard SUM to the last unsharded run. - **2026-08-25 (audit): agent-reported findings need a lead spot-check before filing, and it is cheap.** Six agents returned ~45 candidates; every one the lead re-read at the cited line held, but two summaries overstated a mechanism (an `apply_` stamping claim that needs a `contract.secType` the agent's repro omitted). Reproduce the top P1 in-process from the cited file (a 10-line python heredoc), read the cited lines of every P0/P1, and only then number. - **2026-08-25 (remediate): this runner has no `rtk`, no `setsid`, and no `pytest-xdist`.** Bare `git` is the only git here and its output was trustworthy (the 2026-08-16 rtk lesson applies only where rtk is installed — check `which rtk` first). `pytest-xdist` is CI-only like `pytest-asyncio` was: install it in this loop's venv (`~/radon-weekend/venv-testing` — the legacy shared `~/radon-weekend/venv` is unused since the per-loop split) before verifying anything under `-n auto --dist loadfile` (a new shard is only proven with CI's flags). `setsid` does not exist on darwin: detach a long job with `subprocess.Popen(..., start_new_session=True)`, never `nohup setsid`. - **202 …(truncated)