[noahgift]

armed_shapes is the tool's claim about what it measured. It named shapes that graded nothing, and the aggregate said Pass. Draft per the lane rule.

The defect, reproduced before anything was changed

// crates/aprender-contracts/src/lint/shapes_gate.rs
if report.focus_nodes_n == 0 {
    return Some(ShapesOutcome::NoFocus { shapes_n });
}

The question is asked globally, so it is only ever true for a lone contract. In a directory the other shapes carry the total above zero and an empty shape is invisible. Fixture shapes-one-empty on origin/main:

verdict: Pass   violations: 0   focus_nodes_n: 1
by_shape:         ['empty-shape=0', 'tool-status=1']
armed_shapes:     ['empty-shape', 'tool-status']      <- the claim
not_armed_shapes: []

by_shape had carried empty-shape=0 all along and nothing read it — the same computed but unexposed class as #3597, and the first instance inside the gate that catches the class. The number that falsifies the claim was in the same JSON object as the claim, computed every run.

The shape generalises, and it is worth the next reader's time: asking the vacuity question globally is correct for the singular case and silently wrong for the plural. A lone contract is the empty case and declines properly; a directory is the full case and passes over the same hole. That is the placeholder-is-most-permissive inversion rotated ninety degrees — the same one as #3609, where a GGUF with no vocabulary gets a synthetic <unk> and works while one with a real vocabulary lacking <unk> is refused. The better-specified input fails and the degenerate one passes. Anything that tests a collection's total rather than each member has this available to it.

After

case	verdict	exit	declines
armed shape at zero, in a directory	decline NoFocus	2	named in stderr
unarmed shape at zero	Pass	0	["empty-shape"] — reported, not swallowed
every shape graded something	Pass	0	[]
lone contract that cannot extract	decline	≠0	unchanged

shapes: 1 ARMED shape(s) graded ZERO focus nodes and cannot be counted as clean
        (2 shape(s), 1 focus node(s) in total): empty-shape

The unarmed row is why this is not "refuse whenever anything is empty"

An unarmed shape at zero did not affect the verdict, so it is reported rather than refused — and reported rather than swallowed, which is the half the original defect got wrong. A declines list that could only ever be empty would be decoration, which is the defect one layer up from this one; it carries the unarmed empties so the field is live.

Acceptance, against what shipped

1. zero focus nodes never in armed_shapes ✅ — excluded from the report. Checked first that this is safe for the arming ratchet: check_monotone compares lint-baseline.json declarations, not this report's list. They are put in declines, not not_armed_shapes, because "not armed by policy" and "armed and measured nothing" are different facts and folding the second into the first would file a vacuity as a deliberate choice — which is how it hid.
2. aggregate carries a declines: list ✅
3. --gate shapes refuses when an armed shape graded nothing ✅ exit 2

Both venues, and the mutation

The lone-contract venue already declined correctly, so a fix proved only there proves nothing. The committed table covers both and RED-turns on the defect: disabling the per-shape check fails an_armed_shape_that_graded_nothing_refuses_in_a_directory and the_three_answers_remain_distinct; restoring gives 5/5.

The real corpus is unaffected

pv lint contracts --gate {sigma,relations,shapes} on aprender: Pass, 0 violations, every armed shape with focus nodes (ladder-measured=8, ladder-green=7, ont-shapes-v1=1734), declines: []. The fix refuses the vacuous case and nothing else.

Evidence: evidence/pv/3610/MEASUREMENT.md.

keep-open: #3610's arm 2 (a jsonl extractor, one row per line = one focus node) is NOT in this PR — this is arm 1, the vacuity refusal, which the issue calls "the one that matters" because with it the jsonl workaround is unnecessary. #3567 is the must-carry whose non-vacuity floor this makes real. #3597 is the audit class this is an instance of.

Refs #3610, #3567, #3597

🤖 Generated with Claude Code

[github-actions]

§13.11 rung 1 — quorum shadow verdict

S13-SHADOW pr=3622 head=81460f001b3554af94d965e5ed06eec73b50629e verdict=REFUSE class=Q1 arm_rc=1

Shadow mode: this records a verdict and merges nothing. A refusal
to arm is not a block (§13 adds zero rows to §7) — the pull request is
exactly as green as it was.

[noahgift]

Disclosure first, because the exemption depends on it: I did not run paiml-implement Phase-1 routing on this PR. I authored it directly. So I cannot state that "Phase-1 routing fired no trigger" — there was no routing step to fire. Anything I said to that effect would be a decision reconstructed after the fact, which is exactly what the receipt exists to prevent.

What follows is a footprint measured now, from the PR's own file list, labelled as post-hoc.

#3622 — measured |M| = 2. I read Q1 as NOT firing, and I am not re-arming on my own say-so.

15 files, but 9 are tests/fixtures (data), and ci/explicit-test-commands.d, evidence/pv, scripts/tree_reader_tests.txt are wiring and evidence rather than owning modules. The code modules are:

1. aprender-contracts::lint — mod.rs and shapes_gate.rs, one module, two files
2. aprender-contracts-cli — one integration test target

|M| = 2, which is under the Q1 threshold. The "15 files, 2 crates" count that triggered the disarm is real, but files and crates are not what Q1 measures.

I am stating that as a reading, not acting on it. Your rule, your call — if you want the module count to be a crate count in practice, say so and I will run the lanes. I would rather re-run a quorum I did not need than set a precedent where the author'"s

[noahgift]

Closing keyword: Refs #3610, not Closes — this PR is arm 1 of two

#3610 states two arms and this PR implements one:

arm	this PR
1. Vacuity refusal — any contract whose target resolves to 0 focus nodes renders Unknown{NoFocusNodes} (exit 2), never Pass	done — per-shape, both venues, with the unarmed direction reported in declines[] rather than refused
2. jsonl extractor — one row per line = one focus node	not touched

The issue says "Arm 1 is the one that matters", and I agree — with it, the workaround is unnecessary because the gate says so on day one. But "the arm that matters" is not "the issue's criterion", and a closing keyword claims the latter. Arm 2 is what makes a type: jsonl contract actually gradeable; until it lands, the representative-json-entity workaround and the unmeasured: [jsonl-rows] annotations stay load-bearing.

So: Refs #3610, and the issue stays open on arm 2 after this merges.

[noahgift]

quorum-review (AD-04): NOT agreed (auto_merge: checked=true was_armed=false disarmed=false)

{
 "ticket": "PMAT-3610",
 "head": "7a9a7d22015301c708972f091a06885aad0206a3",
 "width": 3,
 "executor": "agy",
 "agreed": false,
 "auto_merge": {
  "checked": true,
  "was_armed": false,
  "disarmed": false,
  "note": "auto-merge not armed"
 },
 "lanes": [
  {
   "lane": 1,
   "verdict": "FAIL",
   "findings": 3
  },
  {
   "lane": 2,
   "verdict": "FAIL",
   "findings": 3
  },
  {
   "lane": 3,
   "verdict": "PASS",
   "findings": 3
  }
 ]
}

[noahgift]

quorum-review (AD-04): three PASS — agreed (auto_merge: checked=true was_armed=false disarmed=false)

{
 "ticket": "PMAT-3610",
 "head": "1c554897efc605e7bff4bb4ed5a6363e2cc6f9d1",
 "width": 3,
 "executor": "agy",
 "agreed": true,
 "auto_merge": {
  "checked": true,
  "was_armed": false,
  "disarmed": false,
  "note": "auto-merge not armed"
 },
 "lanes": [
  {
   "lane": 1,
   "verdict": "PASS",
   "findings": 5
  },
  {
   "lane": 2,
   "verdict": "PASS",
   "findings": 5
  },
  {
   "lane": 3,
   "verdict": "PASS",
   "findings": 3
  }
 ]
}

[noahgift]

Cross-inspection of docs/audits/quorum-PMAT-3610.json lanes (non-author: aprender-04, 00:03Z)

lane	conversation	status	verdict	findings	refs to sibling lanes / receipts / $WORK	duration
lane 1	10b8b8f5	SUCCESS	PASS (structured_output)	5 (5 cited, 0 with commands)	0	211 s
lane 2	81693b02	SUCCESS	PASS (structured_output)	5 (5 cited, 0 with commands)	0	26 s
lane 3	22217a09	SUCCESS	PASS (structured_output)	3 (2 cited, 1 with commands)	0	130 s

Distinct agy conversation ids: 3/3; no lane references a sibling lane, another receipt or $WORK. Read from /mnt/nvme-raid0/agent-wt/rel-3610/docs/audits/quorum-PMAT-3610.json.lanes on this box. Note: passes after repair — earlier rounds drew real FAILs that were fixed; trio pro-high / pro-low / 3.6-flash, no silences.

Verdict line: 3/3 PASS, independent. Arming (auto-merge waits for the head's checks).

[noahgift]

quorum-review (AD-04): NOT agreed (auto_merge: checked=true was_armed=false disarmed=false)

{
 "ticket": "PMAT-3610",
 "head": "c3ed2e07da949d3b3d0056f48f0c2bdb721cd3ac",
 "width": 3,
 "executor": "agy",
 "agreed": false,
 "auto_merge": {
  "checked": true,
  "was_armed": false,
  "disarmed": false,
  "note": "auto-merge not armed"
 },
 "lanes": [
  {
   "lane": 1,
   "verdict": "FAIL",
   "findings": 3
  },
  {
   "lane": 2,
   "verdict": "PASS",
   "findings": 5
  },
  {
   "lane": 3,
   "verdict": "PASS",
   "findings": 6
  }
 ]
}

[noahgift]

quorum-review (AD-04): three PASS — agreed (auto_merge: checked=true was_armed=false disarmed=false)

{
 "ticket": "PMAT-3610",
 "head": "3aab786b1b0cbf689449faac68b534e4df4d2f0e",
 "width": 3,
 "executor": "agy",
 "agreed": true,
 "auto_merge": {
  "checked": true,
  "was_armed": false,
  "disarmed": false,
  "note": "auto-merge not armed"
 },
 "lanes": [
  {
   "lane": 1,
   "verdict": "PASS",
   "findings": 5
  },
  {
   "lane": 2,
   "verdict": "PASS",
   "findings": 5
  },
  {
   "lane": 3,
   "verdict": "PASS",
   "findings": 5
  }
 ]
}

[noahgift]

Cross-inspection of docs/audits/quorum-PMAT-3610.json lanes, re-review (non-author: aprender-04, 02:03Z)

lane	conversation	status	verdict	findings	refs to sibling lanes / $WORK	duration
lane 1	40cc0669	SUCCESS	PASS (structured_output)	5 (5 cited, 0 with commands)	0	18 s
lane 2	e1af8160	SUCCESS	PASS (structured_output)	5 (5 cited, 0 with commands)	0	16 s
lane 3	f23681f7	SUCCESS	PASS (structured_output)	5 (5 cited, 0 with commands)	0	91 s

Distinct agy conversation ids: 3/3; no lane references a sibling lane or $WORK. Read from /mnt/nvme-raid0/agent-wt/rel-3610/docs/audits/quorum-PMAT-3610.json.lanes on this box. Note: re-review after the large_enum_variant allow (wire-format reason) and the lone-venue assertion; #3648 split out with a self-deleting test.

Verdict line: 3/3 PASS, independent. Re-arming.

[noahgift]

Folded into the 0.69 release batch #3669 (08:25Z) and taken out of the queue so it isn't built twice. Its receipt and closing references are carried in #3669.

[noahgift]

Taken back out of release batch #3669 (08:45Z). The batch's CI traced a red guard to this PR: the complexity ratchet flagged `run_shapes_gate`. The finding persists without this PR (it comes from #3600), but this PR adds to the same function, so it goes to the next batch once #3600's refactor lands. The receipt stands. It rides the next batch once fixed. The owner's session is gone, so this is adoptable by selector.

[noahgift]

Closed under the operator rule of 2026-09-24 ("no PR open over 1 day"), relayed by aprender-49 on the cop's (aprender-cf) instruction. The branch PMAT-3610-vacuous-shapes is kept (head f6ab946a6).

Why it cannot land today:

• The PR is DIRTY (it conflicts with main).
• The present check fails.
• The quorum is split 2 PASS / 1 FAIL, with the failing finding at shapes_gate.rs:246, and no round has agreed.
• It was taken out of batch release(0.69): batch — 15 receipted PRs + fleet-pv andon fix, one CI run, one queue slot #3669 because it adds to run_shapes_gate, which the complexity ratchet flags.

Path forward: the 0.70 fold (aprender-6c) can carry this branch once it is rebased and a quorum round agrees. Closed by aprender-98.