MEFORORG · wshallwshall · Sep 6, 2026 · Sep 5, 2026
@@ -15095,7 +15095,9 @@ measurement from this row's subject and it is named here rather than performed.*
 
 ## 1272. the shell-syntax harness resolves whichever bash PATH orders first, and reports a harness failure as 160 content failures
 
-> 🔢 **Re-scored 2026-08-20 -> P3.** Value **2/10** · Difficulty **1/10** · _fill-in_. The underlying defect is real and unfixed (tests/test_workflow_shell_syntax.py:110, tests/test_dependabot_automerge_guardrails.py:145, tests/test_installed_coord_hooks.py:666 all still use unpinned discovery, and the third resolves sh first), but #1216 is the surviving number for it and its amendment already absorbs this item's unique content, naming the three modules and the 127-versus-2 discrimination layer. The remainder specific to THIS number is closing it with a pointer, a ledger edit. _(previously unscored.)_
+> ✅ **SHIPPED -- verified on main c57903c2c.** All three named sites pin the interpreter on proof, not on a name: `require_bash` from `tests/_bash_resolver.py` at test_workflow_shell_syntax.py:123 and test_dependabot_automerge_guardrails.py:212, and at test_installed_coord_hooks.py:719-726 the `sh` preference must pass `bash_sees` and `bash_preserves_path_order` or fall back to the resolver (PR 672). The 127-versus-2 layer is `explain_returncode` at _bash_resolver.py:231, widened to 126 under this number, so a harness fault names itself. Re-running this item's own control with PATH forced so `shutil.which("bash")` returns the WSL stub, the resolver still returned Git Bash usr/bin and the three modules gave 53 passed, 8 skipped, zero failures against the 19 recorded here. #1216 is the surviving number and is SHIPPED.
+>
+> **Re-scored 2026-08-20 -> P3.** Value **2/10** · Difficulty **1/10** · _fill-in_. The underlying defect is real and unfixed (tests/test_workflow_shell_syntax.py:110, tests/test_dependabot_automerge_guardrails.py:145, tests/test_installed_coord_hooks.py:666 all still use unpinned discovery, and the third resolves sh first), but #1216 is the surviving number for it and its amendment already absorbs this item's unique content, naming the three modules and the 127-versus-2 discrimination layer. The remainder specific to THIS number is closing it with a pointer, a ledger edit. _(previously unscored.)_
 >
 > **Filed 2026-08-14 - not started. TWO SEATS RAN THE SAME TEST ON THE SAME BOX ON THE SAME DAY AND GOT OPPOSITE RESULTS.** `tests/test_workflow_shell_syntax.py` writes each workflow shell block to a temp file and checks it with `subprocess.run([bash, "-n", str(probe)])`, where `bash` comes from **`shutil.which("bash")`** and `str(probe)` is a **Windows** path. **Both a Git Bash and a WSL `bash` are installed on this box**, and `shutil.which` returns **whichever the invoking process's `PATH` happens to order first**. Nothing pins it.
 > **MEASURED, BOTH RESOLUTIONS, SAME WINDOWS PATH, WITH A REAL-SYNTAX-ERROR NEGATIVE CONTROL:**
@@ -15457,7 +15459,9 @@ can finish a suite rather than merely a lane with hours.
 
 ## 1291. Get-HandledTools returns NULL on its zero-tool and missing-file paths, correct only by its caller's grace
 
-> 🔢 **Re-scored 2026-08-20 -> P3.** Value **2/10** · Difficulty **2/10** · _fill-in_. Unchanged: both exit paths at install-gate.ps1:244 and :250 still unroll, and the sole caller at :302 still re-wraps, so shipped behaviour is correct and the trap is latent exactly as filed. Value 2 is rung 2 -- developer tooling with no product, PHI or live-defect axis, and the one caller already covers it. Difficulty 2 for the two-token fix plus a test that must reach the function through ParseFile and FunctionDefinitionAst, since dot-sourcing install-gate.ps1 performs a machine-global install and a test routed through -Status cannot tell fixed from unfixed. _(previously unscored.)_
+> ✅ **SHIPPED -- verified on main c57903c2c.** PR #577 comma-wrapped both exit paths of `Get-HandledTools` (scripts/worktree/install-gate.ps1:250 and :256) and dropped the sole caller's compensating `@()` wrap at :308, so the array guarantee now lives in the function rather than in its caller's grace. The discriminating AST-lift test is tests/test_gate_rule_scan_agreement.py:322, asserting type and count on the missing-file, zero-tool, one-tool and two-tool arms without wrapping the result; 26 tests pass, and a scratch copy with either comma removed reproduces the filed NULL / bare-String table.
+>
+> **Re-scored 2026-08-20 -> P3.** Value **2/10** · Difficulty **2/10** · _fill-in_. Unchanged: both exit paths at install-gate.ps1:244 and :250 still unroll, and the sole caller at :302 still re-wraps, so shipped behaviour is correct and the trap is latent exactly as filed. Value 2 is rung 2 -- developer tooling with no product, PHI or live-defect axis, and the one caller already covers it. Difficulty 2 for the two-token fix plus a test that must reach the function through ParseFile and FunctionDefinitionAst, since dot-sourcing install-gate.ps1 performs a machine-global install and a test routed through -Status cannot tell fixed from unfixed. _(previously unscored.)_
 >
 > **Filed 2026-08-19 -- not started. THIS IS A LATENT TRAP, NOT A LIVE DEFECT, AND THE DISTINCTION IS THE ITEM.** `Get-HandledTools` in [`scripts/worktree/install-gate.ps1`](../scripts/worktree/install-gate.ps1) returns `$null` rather than an empty array on two of its paths. **Shipped behaviour is correct today** because its single caller (around line 294) wraps the result in `@(...)`. **A commit message or item claiming a present-tense bug here would be false.** The honest claim: the function is correct only by its caller's grace, and the guarantee belongs in the function. **THE CHANGE: make both exit paths return an array unconditionally.**
 > **MEASURED, in pwsh 7, by extracting the function through the PowerShell AST rather than running it:**
@@ -17988,7 +17992,9 @@ at which point the population is gone and this row cannot be re-derived.
 
 ## 1345. The claim gate decides documentation by path prefix, so executable code under docs escapes it
 
-> 🔢 **Filed 2026-08-23 - not started.** `claim_check.py` decides whether a commit "touches code" by testing the PATH PREFIX against a documentation list. ***A `.py` file under a documentation prefix therefore DECLARES ITSELF DOCUMENTATION TO THE CLAIM GATE***, and a commit touching only such files is stood down on without ever meeting the claim rule.
+> ✅ **SHIPPED -- verified on main c57903c2c.** The claim gate now decides on what a file IS, not where it sits: the code-extension test runs before the location test and is casefolded (scripts/hooks/claim_check.py:175-192, _CODE_SUFFIXES at :73), landed by PR 563 and PR 576. Driving _touches_code returns CODE for docs/bench/microbench.py, docs/security/asvs-scorecard.py, docs/bench/Tool.PY and .github/workflows/ci.yml, and DOC for docs/BACKLOG.md and docs/bench/results.csv, so the deliberate documentation carve-out survives. Both controls are pinned by paired arms plus an anti-vacuity check on the live population (tests/test_claim_check.py:223, :233, :242 must-block; :264, :274 must-not-block; :284), 18 passed.
+>
+> **Filed 2026-08-23 - not started.** `claim_check.py` decides whether a commit "touches code" by testing the PATH PREFIX against a documentation list. ***A `.py` file under a documentation prefix therefore DECLARES ITSELF DOCUMENTATION TO THE CLAIM GATE***, and a commit touching only such files is stood down on without ever meeting the claim rule.
 >
 > **Scored 2026-09-03 -> P3.** Value **1/10** · Difficulty **1/10** · _fill-in_. Both required controls are built and live, and I drove them rather than read them. The extension test now runs BEFORE the location test and is casefolded (scripts/hooks/claim_check.py:148-166, list at :57), so importing the module and calling _touches_code returned CODE for docs/bench/microbench.py, docs/security/scorecard.py, docs/bench/Tool.PY and .github/workflows/ci.yml, and DOC for docs/BACKLOG.md and docs/bench/results.csv -- the exempt-doc twin the item says must survive. The paired arms exist and assert the right direction (tests/test_claim_check.py:233 must-block, :264 must-not-block, :284 anti-vacuity), and the whole file ran green: 18 passed via pytest tests/test_claim_check.py. The measured population is covered -- 2 .py under docs/ here and 9 files under the vault's doc prefixes (8 .py including the six docs/security/asvs-*.py that write the security record, plus one .yml) all carry listed extensions -- and the fix is on the live path, since the installed .git/hooks/claim_check.py carries _CODE_SUFFIXES and the casefolded test, wired through the commit-msg hook. Only bookkeeping is left: docs/BACKLOG.md:16228 still banners this "not started", a banner flip and nothing runnable.
 > Verdict: build
@@ -19034,7 +19040,9 @@ share a shape.
 
 ## 1367. The gate-parity failure asks the reader to attribute the drift instead of computing it
 
-> 🔢 **Filed 2026-08-26 -- FIXED IN THIS CHANGE; the status banner is deliberately unchanged**
+> ✅ **SHIPPED -- verified on main c57903c2c.** PR 632 added the computed attribution strictly additively (+186/-0): the pure `drift_verdict` at tests/test_gate_installed_parity.py:317 holds all four branches the item's table names (installed==main at :342, checkout==main at :359, neither at :365, main unreadable at :337), and the untouched assertion interpolates it at :432. Driven at HEAD, the four branches return distinct and correctly directed text, and the file runs 13 passed -- covering the four branch tests (:668, :687, :694, :699), the mutual-distinctness control (:715), and the wiring test that monkeypatches INSTALLED_GATE to a temporary copy rather than the machine-global file (:727). No doc or record limb is left: docs/BACKLOG.md:282 already ranks this row confirmed shipped.
+>
+> **Filed 2026-08-26 -- FIXED IN THIS CHANGE; the status banner is deliberately unchanged**
 > (same handling as #1026, #1361 and #1365). `test_the_installed_gate_matches_the_committed_source`
 > compares the installed gate against THIS CHECKOUT, which is the right question. Its failure text
 > then correctly REFUSES to name a culprit and hands the reader a `git log` to run. Three seats did
@@ -19107,7 +19115,9 @@ take it out from under a concurrent session. Same reasoning the existing negativ
 
 ## 1366. the four connscale fields that discriminate the surviving failure hypotheses never escape the test job
 
-> 🔢 **Filed 2026-08-26 (builder 2) - BUILT IN THIS COMMIT, not yet landed.** Successor in subject to [#1211](#1211-empty_claims_per_msg-is-not-contention-immune-the-ratio-form-excursions-past-its-own-slo-band-on-a-hosted-runner), whose limb one made the RATIO survive a passing run; this makes the readings that EXPLAIN the ratio survive one too.
+> ✅ **SHIPPED -- verified on main c57903c2c.** PR 624 landed all five band-less fields as DIAGNOSTIC_FIELDS (harness/load/connscale/report.py:597) plus the second renderer render_diagnostics_markdown (report.py:907), whose table carries no prior, band floor, margin or verdict and whose preamble states that none of them has an SLO band. Driven rather than read: a record with reload_seconds=None and fd_probe_degraded_ticks=0 renders the unmeasured field as a dash and the measured zero as 0, the fixture calls _record_diagnostics unconditionally at tests/test_connscale_smoke.py:450, and tests/test_connscale_empty_claims_per_msg.py gives 47 passed. The 2026-08-31 correction is carried in the code at report.py:918, which names fd_count_monotonic as the only enforced band.
+>
+> **Filed 2026-08-26 (builder 2) - BUILT IN THIS COMMIT, not yet landed.** Successor in subject to [#1211](#1211-empty_claims_per_msg-is-not-contention-immune-the-ratio-form-excursions-past-its-own-slo-band-on-a-hosted-runner), whose limb one made the RATIO survive a passing run; this makes the readings that EXPLAIN the ratio survive one too.
 >
 > **Scored 2026-09-03 -> P3.** Value **1/10** · Difficulty **1/10** · _fill-in_. Verified by driving the code, not by reading it, and every limb is met. All five band-less fields the item enumerates are in DIAGNOSTIC_FIELDS at harness/load/connscale/report.py:594-621, and the second renderer render_diagnostics_markdown at harness/load/connscale/report.py:904 emits them with no prior, band floor, margin or verdict column plus a preamble sentence stating none has an SLO band -- I imported the module and rendered a record with reload_seconds=None and fd_probe_degraded_ticks=0, which produced "| 1.25 | - | 7 | 0 | 2.5 |", so an unmeasured field is a dash and a measured zero still prints 0. The fixture calls _record_diagnostics unconditionally before any assertion at tests/test_connscale_smoke.py:450, routed through _append_step_summary at tests/test_connscale_smoke.py:271, which is the selection-bias limb. The test arm is real and collected: tests/test_connscale_empty_claims_per_msg.py:1117-1225 pins the shipped set against a literal rather than against DIAGNOSTIC_FIELDS itself and asserts the unconditional call site by source scan, and running that file on the repo venv gave 47 passed. The item's own 2026-08-31 correction is carried into the code, with harness/load/connscale/report.py:915 naming fd_count_monotonic as the only enforced band, so no docstring still reads the retired "two metrics" way.
 > Verdict: build
@@ -21231,7 +21241,9 @@ The mark is applied at runtime by `tests/conftest.py` from that manifest, **not
 
 ## 1410. gate liveness treats a timeout-cancelled job as did-not-run, so a gate killed mid-measurement reports healthy
 
-> 🔢 **Filed 2026-08-31 (ci-diagnostics) - BUILT IN THIS COMMIT, not yet landed.** The reason #1409 ran eight days unseen. Fixed in the same change.
+> ✅ **SHIPPED -- verified on main c57903c2c.** PR 724 (merge 4107615cc) keyed the did-not-run excuse on the receipt being ABSENT instead of on the result string (scripts/quality/liveness.py:207-210) and joined `cancelled` to the anti-laundering branch with the reworded "A gate that did not finish is dead, not inapplicable" (liveness.py:285-300). Replaying the four-PR production payload against both file versions confirms the direction: the pre-fix file returns no violations, HEAD returns "the gate reported ITSELF dead", and a genuine cancel with no receipt stays green in both. `pytest tests/test_gate_liveness.py -q` gives 31 passed, covering the must-trip, must-not-trip and laundering arms (tests/test_gate_liveness.py:246, :260, :285); no `negative_controls.toml` entry exists, as intended.
+>
+> **Filed 2026-08-31 (ci-diagnostics) - BUILT IN THIS COMMIT, not yet landed.** The reason #1409 ran eight days unseen. Fixed in the same change.
 >
 > **Scored 2026-09-03 -> P3.** Value **1/10** · Difficulty **1/10** · _fill-in_. Every limb was checked by running it, not by reading it. The receipt-conditional excuse is live at scripts/quality/liveness.py:206-257 and the workflow really calls it (.github/workflows/quality-advisory.yml:802-803, with the coverage job's `if: always()` record step and `receipt` output at :510-512 and :341-342); loading the pre-fix file from 4107615cc^ beside HEAD's and feeding both the four-PR production payload, the old one returns no violations while the new one returns "the gate reported ITSELF dead", and a genuine cancel with no receipt stays green in both, so the must-trip and must-not-trip arms are confirmed by direction. The anti-laundering rider is at liveness.py:285-304 with the reworded "A gate that did not finish is dead, not inapplicable", and `pytest tests/test_gate_liveness.py -q` gives 31 passed using the primary checkout's .venv, which closes the gap the item itself named as unrun; no negative_controls.toml entry exists for `gate liveness (advisory)`, as intended. All that remains is flipping the still-open banner at docs/BACKLOG.md:18855, a doc edit that ships nothing runnable. Note for a future filer, not a defect of this item: the pre-existing signal-field hole at liveness.py:147 is explicitly scoped out here and still has no number of its own.
 > Verdict: build