Research: certify a class-C V2 stabilizer quotient

research/k2-class-c-stabilizer-orbit-quotient/NOTICE.md

@@ -0,0 +1,5 @@
+# Third-party input notice
+
+`research/k2-class-c-stabilizer-orbit-inputs/data88.txt` is copied unchanged from DreamingOfClouds' repository package [`research/k2-class-c-help`](https://github.com/DreamingOfClouds/integer-mult-bounds/tree/research/k2-class-c-help/research/k2-class-c-help), commit `1b25759bff5486293a9083ebbe0097b5f6642e08`. It is part of that Apache-2.0 package, whose README attributes preparation to DreamingOfClouds with Claude Opus 5.5 assistance. SHA256: `7a207989d901f488120988007dcad3d8edf3b8b1ba79eac7307256925b915667`.
+
+The new verifier and proof text in this directory are contributed under this repository's Apache-2.0 license. No endorsement by the upstream authors is implied.

research/k2-class-c-stabilizer-orbit-quotient/README.md

@@ -0,0 +1,58 @@
+# Class-C V2 stabilizer orbit quotient
+
+**Status: scoped partial result.** This is a sound candidate reduction for one first-flat orbit in the open class-C four-flat checker request. It does not kill a residue row, complete an `s4z` run, or change the selected multiplication result.
+
+## Result
+
+For the config `6554 = (6,5,5,4)` in [Discussion #2954](https://github.com/CrocSwap/integer-mult-bounds/discussions/2954), use the `s4z` ordering `(a,b,c,d)=(5,6,4,5)`. Let `P` be the 112 pinned labels in `F_2^9`; for a subspace `V`, let `Lambda(V)` be the set of P112 indices whose labels lie in `V`. Let `G` be the group generated by the 12 permutations in the pinned `gens.txt`.
+
+The verifier establishes, by exact finite checks, that:
+
+1. Each generator is induced by an invertible linear map on all of `F_2^9`, checked on the 112 labels and all 512 vectors.
+2. Each generator permutes all 18,100 typed rows of the pinned `data88.txt` catalog by the rule `Off(h) -> g(Off(h))`.
+3. For each such permutation, `REAL_MAPS.json` supplies one exact invertible rational `8 x 8` linear map on `Q^8` carrying every listed one-dimensional or two-dimensional direction space to its catalog image. The verifier checks each matrix exactly against every row; the matrices are finite proof witnesses, not floating-point evidence.
+4. The five-flat with RREF basis `(4,8,32,210,321)` is fixed by the first five generators. Its full G-orbit contains 2,688 distinct root masks; conjugating the subgroup to the least root-mask representative (mask `197427`) gives a subgroup of that representative's stabilizer.
+5. Among all six-dimensional subspaces of `F_2^9`, the 788,035 RREF subspaces induce 776,806 distinct P112 root masks. The conjugated five-generator subgroup partitions those masks into exactly 47,850 orbits. One V2 representative per orbit reduces this first-flat candidate list by the exact factor `776806/47850 = 16.234190177638x`; the largest orbit has 32 masks.
+
+The flat index masks in `s4z` are these intersections with P112, so duplicate six-flats with the same root mask are deliberately counted once, as in the checker.
+
+## Why the quotient is sound
+
+For any subgroup `H <= G` fixing a chosen first flat `V1`, the map `V2 -> g(V2)` for `g in H` is a bijection on the V2 mask universe. The checked label action gives `Lambda(gV)=g Lambda(V)`. The checked direction action maps every tested `Off(h)` to the corresponding support, and the checked invertible rational map preserves real span dimension. Therefore the set of qualifying directions, the feasibility of the remaining two-flat covers, and the final span rank are invariant under `H`. Every V2 orbit has the same result, so retaining one representative per orbit cannot change the checker verdict.
+
+The numerical orbit count is a source-bound exhaustive enumeration. The safety proof for applying the quotient is the equivariance argument above plus the finite checks that bind `G` to the listed catalog. The claim is conditional on the upstream statement that `data88.txt` is complete for directions with `|Off(h)| <= 88`.
+
+## Scope and limits
+
+- The count applies to one G-orbit of five-flats, for the `(5,6,4,5)` ordering of config `6554`. It does not give the aggregate reduction across all first-flat orbits.
+- The verifier uses the least root-mask representative convention to make the conjugating witness deterministic. The pinned `s4z.c` uses union-find roots instead; deriving the conjugate subgroup for each existing representative or adopting the least-mask convention is integration work.
+- No `s4z` pair closure was run. The 16.234x number is a candidate-count reduction for the V2 mask set, not a wall-clock speedup.
+- No row in `rows17.csv` is killed. No part of the conditional RRC result is verified here.
+- No supplier, lower bound for all multiplication algorithms, or `kappa` improvement is claimed. The reviewed selected result remains `kappa = 0.001672660197076918` in the scoped, conditional construction.
+
+## Reproduce
+
+From the repository root, after the package and its inputs are committed:
+
+```sh
+python3 -B research/k2-class-c-stabilizer-orbit-quotient/verify.py
+python3 scripts/research_package.py --package research/k2-class-c-stabilizer-orbit-quotient --check-only
+python3 scripts/research_package.py --package research/k2-class-c-stabilizer-orbit-quotient
+```
+
+The verifier uses Python's standard library, exact integers, and `fractions.Fraction`. It checks the three declared source hashes and includes rejecting controls for a non-permutation, a bad support action, and a singular real-space map. The registered replay allows at most 600 seconds.
+
+## Sources and related work
+
+- Discussion #2954 defines the Class-C question and the 6554 configuration.
+- `data88.txt` is copied unchanged from DreamingOfClouds' Apache-2.0 package at [`research/k2-class-c-help`](https://github.com/DreamingOfClouds/integer-mult-bounds/tree/research/k2-class-c-help/research/k2-class-c-help), source commit `1b25759bff5486293a9083ebbe0097b5f6642e08`. Its SHA256 is recorded in `SOURCES.json` and rechecked by the verifier. See [NOTICE.md](NOTICE.md) for attribution.
+- The label order and 12 permutations are pinned to canonical main's `research/one-sided-excess-three/data/data80.txt` and `gens.txt`.
+- PRs [#2975](https://github.com/CrocSwap/integer-mult-bounds/pull/2975) and [#3000](https://github.com/CrocSwap/integer-mult-bounds/pull/3000) optimize requirement enumeration and maximal-mask filtering, respectively. Merged Merged PR [#3315](https://github.com/CrocSwap/integer-mult-bounds/pull/3315) audits which published configs cover the hard rows. PR [#3334](https://github.com/CrocSwap/integer-mult-bounds/pull/3334) supplies the threshold-preserving dominance reduction to `6554@T7` for the hard rows, but no geometric span pass. This subgroup quotient addresses the separate V2 orbit reduction step.
+
+## Open next questions
+
+See the three structured questions in [TASK.json](TASK.json): integrate per-V1 stabilizers into the pinned checker, measure the total effect across nonredundant configs, and independently regenerate the bounded direction catalog.
+
+## Credit
+
+The `data88.txt` input and its E8/P112 direction enumeration are DreamingOfClouds' work, prepared with Claude Opus 5.5 assistance, under the upstream Apache-2.0 license. New proof text and verifier in this package were prepared by OpenAI Codex with substantial AI assistance; no independent human review is claimed. The existing P112 and residue-closure lineage remains credited to its original authors.

research/k2-class-c-stabilizer-orbit-quotient/REAL_MAPS.json

@@ -0,0 +1 @@
+{"schema_version":1,"source_data88_sha256":"7a207989d901f488120988007dcad3d8edf3b8b1ba79eac7307256925b915667","source_gens_sha256":"0679ba17f68761c9aea31230df8a7ac4a86533b24ee53d786dd1afa216a4b2c9","maps":[[[[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[1,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[1,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[1,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[1,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[1,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[1,1]]],[[[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1]]],[[[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[1,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[1,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1]]],[[[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1]]],[[[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[1,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[1,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1]]],[[[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1]]],[[[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[1,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[1,1],[0,1]]],[[[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1]]],[[[-1,2],[1,2],[0,1],[0,1],[-1,2],[1,2],[0,1],[0,1]],[[-1,2],[1,2],[0,1],[0,1],[1,2],[-1,2],[0,1],[0,1]],[[0,1],[0,1],[-1,2],[1,2],[0,1],[0,1],[-1,2],[1,2]],[[0,1],[0,1],[-1,2],[1,2],[0,1],[0,1],[1,2],[-1,2]],[[0,1],[0,1],[-1,2],[-1,2],[0,1],[0,1],[-1,2],[-1,2]],[[0,1],[0,1],[1,2],[1,2],[0,1],[0,1],[-1,2],[-1,2]],[[-1,2],[-1,2],[0,1],[0,1],[-1,2],[-1,2],[0,1],[0,1]],[[1,2],[1,2],[0,1],[0,1],[-1,2],[-1,2],[0,1],[0,1]]],[[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1]],[[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[1,1],[0,1],[0,1],[0,1],[0,1]],[[-1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[1,1],[0,1],[0,1],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[0,1],[-1,1],[0,1],[0,1]]],[[[-1,2],[-1,2],[0,1],[0,1],[0,1],[0,1],[-1,2],[-1,2]],[[-1,2],[-1,2],[0,1],[0,1],[0,1],[0,1],[1,2],[1,2]],[[0,1],[0,1],[-1,2],[1,2],[-1,2],[1,2],[0,1],[0,1]],[[0,1],[0,1],[-1,2],[1,2],[1,2],[-1,2],[0,1],[0,1]],[[0,1],[0,1],[-1,2],[-1,2],[-1,2],[-1,2],[0,1],[0,1]],[[0,1],[0,1],[-1,2],[-1,2],[1,2],[1,2],[0,1],[0,1]],[[-1,2],[1,2],[0,1],[0,1],[0,1],[0,1],[-1,2],[1,2]],[[-1,2],[1,2],[0,1],[0,1],[0,1],[0,1],[1,2],[-1,2]]],[[[-1,2],[-1,2],[-1,2],[1,2],[0,1],[0,1],[0,1],[0,1]],[[1,2],[1,2],[-1,2],[1,2],[0,1],[0,1],[0,1],[0,1]],[[0,1],[0,1],[0,1],[0,1],[-1,2],[1,2],[-1,2],[-1,2]],[[0,1],[0,1],[0,1],[0,1],[1,2],[-1,2],[-1,2],[-1,2]],[[0,1],[0,1],[0,1],[0,1],[-1,2],[-1,2],[-1,2],[1,2]],[[0,1],[0,1],[0,1],[0,1],[-1,2],[-1,2],[1,2],[-1,2]],[[-1,2],[1,2],[-1,2],[-1,2],[0,1],[0,1],[0,1],[0,1]],[[-1,2],[1,2],[1,2],[1,2],[0,1],[0,1],[0,1],[0,1]]]]}

research/k2-class-c-stabilizer-orbit-quotient/SOURCES.json

@@ -0,0 +1,36 @@
+{
+  "schema_version": 1,
+  "base_commit": "c1c5c0c6a897c23686957caae19105a149eec091",
+  "files": [
+    {
+      "path": "certificates/selected-result.json",
+      "usage": "provenance",
+      "git_blob": "b1fea0c94436b236cc2b7d41a7be032811b94c81",
+      "sha256": "bb2eaa58647b353f88d0aec2eafb9d30df1be53589231b23e12ceef3e4bad129"
+    },
+    {
+      "path": "docs/research/agent-task-catalog.json",
+      "usage": "provenance",
+      "git_blob": "67f085acd86acd7d4e03922d9cf4b26bca66fed5",
+      "sha256": "47b42433f5a85809a3aadd29c2c339da2c6ed728e4f7762df6b88bccb7874cf0"
+    },
+    {
+      "path": "research/k2-class-c-stabilizer-orbit-inputs/data88.txt",
+      "usage": "input",
+      "git_blob": "e399477b459075d459a6cde0f4afe040501f0419",
+      "sha256": "7a207989d901f488120988007dcad3d8edf3b8b1ba79eac7307256925b915667"
+    },
+    {
+      "path": "research/one-sided-excess-three/data/data80.txt",
+      "usage": "input",
+      "git_blob": "80f3870046400588c45d8ac32fb131dd34736157",
+      "sha256": "d23ad440471c3afd1a894b3439e29a05f498edfe2af33d7a667ca1d8f1d6b9c5"
+    },
+    {
+      "path": "research/one-sided-excess-three/data/gens.txt",
+      "usage": "input",
+      "git_blob": "ba27ea667b9d00e6976c11d349fa84d959b78253",
+      "sha256": "0679ba17f68761c9aea31230df8a7ac4a86533b24ee53d786dd1afa216a4b2c9"
+    }
+  ]
+}

research/k2-class-c-stabilizer-orbit-quotient/TASK.json

@@ -0,0 +1,73 @@
+{
+  "schema_version": 1,
+  "task_id": "replay-runtime",
+  "priority_revision": 6,
+  "base_commit": "c1c5c0c6a897c23686957caae19105a149eec091",
+  "original_card": {
+    "id": "replay-runtime",
+    "question": "Can one measured replay bottleneck be sped up without dropping coverage?",
+    "starting_points": [
+      "scripts/run_isolated_tests.py",
+      "scripts/verify_replay_coverage.py",
+      "certificates/selected-result.json"
+    ],
+    "methods": [
+      "Profile one isolated replay and optimize the measured bottleneck.",
+      "Design source-bound reuse with explicit dependency invalidation and negative controls."
+    ],
+    "acceptance_gate": "Report before/after wall time on the same fixture, retained checks and invalidation behavior.",
+    "first_step": {
+      "command": [
+        "python3",
+        "-B",
+        "scripts/research_baseline.py"
+      ],
+      "expected": "PASS_RESEARCH_BASELINE_BINDINGS_ONLY",
+      "inspect": "Read the selected reproduction command and choose a single replay stage for profiling.",
+      "scope": "Positive control / source-bound orientation only; this is not a new construction or full cold replay."
+    },
+    "first_experiment": "Measure one unchanged fixture, preserve check coverage, and test one optimization with invalidation controls.",
+    "bottleneck_link": "Faster trustworthy replays increase review throughput and let mechanism candidates be checked sooner.",
+    "allowed_variations": [
+      "Profile one stage or design one source-bound cache boundary."
+    ],
+    "scope_change_triggers": [
+      "Skipping coverage or comparing different fixtures to claim speedup."
+    ],
+    "overlap_terms": [
+      "replay runtime",
+      "cache invalidation",
+      "isolated tests"
+    ]
+  },
+  "scope_status": "changed",
+  "actual_scope": "For the live #2954 class-C 6554 four-flat configuration, can a provably sound stabilizer subgroup quotient the six-dimensional V2 mask enumeration for one five-dimensional V1 orbit?",
+  "scope_change_reason": "The task catalog has no card for the active #2954 route-B checker handoff. replay-runtime is the closest method card, but its original scope concerns selected-result replays. This changed investigation certifies a finite symmetry quotient for the separately pinned s4z direction catalog; it does not claim an end-to-end s4z speedup or a class-C row exclusion.",
+  "bottleneck_link": "The #2954 route-B handoff identifies V2 pair enumeration as a major cost before four-flat closure. A sound stabilizer quotient can reduce candidate V2 masks while preserving the exact span predicate.",
+  "budget_minutes": 60,
+  "stop_rule": "Stop after an exact group-action proof and a fully enumerated bounded V1/V2 mask orbit slice; do not infer any row kill, full-config pass, or kappa change without integrating and completing the s4z closure search.",
+  "outcome": "partial",
+  "evidence": "The verifier pins the 18,100-row data88 catalog, confirms all 12 generators act linearly and invertibly on F2^9, checks each induced Off-support permutation and exact invertible Q^8 matrix witness on every listed line/plane direction, then exhaustively partitions the 776,806 unique six-flat root masks into 47,850 orbits under a five-generator subgroup fixing a canonical five-flat mask. This is one first-flat group orbit and a candidate-count result only; catalog completeness is inherited from the pinned source.",
+  "author": "OpenAI Codex",
+  "ai_assistance": "Prepared and verified by OpenAI Codex with substantial AI assistance. No independent human proof review is claimed.",
+  "open_questions": [
+    {
+      "question": "Can the subgroup quotient be integrated into the pinned s4z engine for every first-flat orbit of config 6554?",
+      "why_it_matters": "The measured 16.234x candidate reduction is for one first-flat orbit; it does not establish the aggregate reduction or preserve the existing run output after code integration.",
+      "first_step": "Use Schreier witnesses to generate stabilizer subgroups for each canonical V1 mask, then compare exact per-pair candidate sets against unchanged s4z on the bounded existing fixture.",
+      "blocked_by": "The pinned s4z source is external to canonical main and uses its own union-find representative order; integration and same-input baseline replay remain to be done."
+    },
+    {
+      "question": "What aggregate time and memory change results across all nonredundant class-C configurations and the 17 rows?",
+      "why_it_matters": "Candidate-count reduction alone does not determine end-to-end closure cost or whether any row is killed.",
+      "first_step": "After the config audit in merged PR #3315 and the threshold-preserving reduction in PR #3334, run one time-capped 6554@T7 shard with the quotient and its unchanged control, then account for all remaining shards before extrapolating.",
+      "blocked_by": "Full per-V1 stabilizer preparation and the expensive closure searches remain unmeasured."
+    },
+    {
+      "question": "Is the 18,100-row data88 direction catalog complete for every direction with |Off(h)| <= 88?",
+      "why_it_matters": "The symmetry check is exact on the pinned rows but cannot certify omitted directions.",
+      "first_step": "Independently regenerate the bounded direction catalog from the E8 frame split and compare normalized supports and subspaces.",
+      "blocked_by": "The source package states completeness, but this contribution does not regenerate its direction enumeration."
+    }
+  ]
+}

research/k2-class-c-stabilizer-orbit-quotient/VERIFY.json

@@ -0,0 +1,22 @@
+{
+  "schema_version": 1,
+  "status": "ready",
+  "python": "3.12",
+  "requirements": [],
+  "inputs": [
+    "research/k2-class-c-stabilizer-orbit-inputs/data88.txt",
+    "research/one-sided-excess-three/data/data80.txt",
+    "research/one-sided-excess-three/data/gens.txt"
+  ],
+  "commands": [
+    {
+      "argv": [
+        "{python}",
+        "-B",
+        "{package}/verify.py"
+      ],
+      "timeout_seconds": 600,
+      "checks": "Source SHA256 pins and label-order identity; all 12 invertible F2^9 actions; all 18,100 typed Off-support permutations and exact Q^8 line/plane maps; exact RREF/Gaussian counts; full V2 mask orbit partition; malformed-permutation, bad-support, and singular-map rejection controls."
+    }
+  ]
+}