trading: rc6 gate rulings: G3 qualified by the retained first replay; G4 excluded (dated)

[seathatflowsinourveins]

Scope

• What this PR changes: one dated record, blueprints/us-equities/engine-nautilus/rc6-gate-rulings-20261007.json, with the gate owner's rulings on the NautilusTrader rc6 candidate's G3 and G4 (the grand-catalog lane's question Q26).
• Why: the command center ruled on both. G3 is qualified by the retained first replay. G4 becomes a dated exclusion rather than an open-ended gate.
• Base commit: origin/main at branch time.
• Lane: lane:trading.
• Owned paths touched: that one file. It needs no registry entry, so manifests/evidence.json and manifests/stack.json are untouched.

The rulings:

• G3: (a). The retained first replay first-rc6-spy-20261006T055742Z (NativeStack2604, one_zero, run and compare exit 0 and 0, 136 PASS with a 106-check execution subset that is never added to it) is G3's qualifying result, with its original date and pre-C5 scope. It covers exactly the seven reviewed rc6 files at the hashes in deviation-acceptance-rc6.json. If any of them changes, a new qualification record comes before any new qualifying run.
• G4: EXCLUDED, dated 2026-10-07. The AAPL baseline and fee_slippage_stress cases need the retained Alpaca run that equity-replay/receipt.json identifies only by hash (a59c6ed7… for the receipt, adb2a240… for the selected rows). A search on NativeStack found neither hash: 7,785 Alpaca-path JSON files and 46,980 receipt or AAPL-named files hashed, no match. NativeStack2604 holds no trading data by design.
  • Overturn route: a dated deviation for a hash-gated reacquisition of the 15 sessions 2020-08-10..2020-08-28, qualifying only if the selected rows hash to adb2a240…. It belongs to the trading lane's next data wave.
  • Standing note: the Mac host's lane searches for the two hashes when that host is next online.
  • No rc6 launcher vector is written without the inputs. rc6's other checks proceed.

SOTA sources

• The rc6 candidate, PR Qualify the isolated NautilusTrader rc6 runtime candidate #761 at 577101588db61764773bdb1651d60e88eeb24025:
  • blueprints/us-equities/engine-nautilus/spy-parity/deviation-acceptance-rc6.json (the first-only seal and the seven reviewed hashes);
  • evidence/receipts/nautilus-rc6-spy-first-replication-20261006.json (the first replay's run block and result);
  • blueprints/us-equities/engine-nautilus/equity-replay/receipt.json (the AAPL input hashes and command template);
  • docs/decisions/2026-10-06-nautilus-rc6-runtime-candidate.md (AAPL blocked).
• The command center's items task-native-agent-stack-5f-20261006T194228Z and task-native-agent-stack-5f-20261007T003642Z.

Evidence-class table

Claim	Evidence class	Command / receipt
G3's result and scope	repository record at a pinned commit	the first-replication receipt and the acceptance seal at 577101588
G4's inputs are on no searched path of NativeStack	local host search (absence within the stated scope only)	SHA-256 of the files named in the record's search block, no match for either hash

Local commands run

$ python3 -c "import json; json.load(open('blueprints/us-equities/engine-nautilus/rc6-gate-rulings-20261007.json'))"
$ python3 scripts/validate.py
{"components": 69, "hashed_files": 10388, "profiles": 4, "receipts": 223, "status": "passed"}

Decision record

The command center's items cited above; this record transcribes their rulings.

Host evidence

Not applicable: this PR adds or changes no file under evidence/hosts/.

• python3 scripts/host_receipts.py validate passes for every new/changed receipt (none; scripts/validate.py exits 0 at 5bdbd64).
• Independent review is recorded or requested: the cross-family connector review threads (P1, P2) are answered and resolved, and the co-op re-reads at this head.
• No platform_status change is made from a host receipt alone (none changes).

Checklist

• New/changed GitHub Actions are pinned to a full commit SHA with a version comment (no workflow changes here).
• New/changed workflows declare top-level permissions: {} and grant each job only what it needs (no workflow changes here).
• No secrets are printed, logged or committed; no new required secret was added.
• No new paid hosting, subscription or billing surface was introduced.
• Peer-owned untracked files and worktrees were preserved.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-07T00:44:18.297742Z	a365052	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a36505214c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[seathatflowsinourveins]

Mac host search for the two G4 AAPL input hashes (Mac coordinator, 2026-10-08 00:41Z): no match.

This is the search rulings.G4.standing_note asks for. Read-only whole-file SHA256 over every regular data-like file. Symlinks were not followed, files were capped at 2 GiB (none exceeded it), and there were 0 read errors. Extensions: csv, tsv, parquet, json, jsonl, gz, zip, zst, feather, arrow, txt, bin, dbn, pkl, npz, h5/hdf5, db and sqlite. node_modules, .git, virtual environments and caches were skipped.

Pass	Roots	Files hashed	Matches
1 (00:39:04-00:39:41Z)	~/Documents/Trading, ~/code, ~/Downloads, ~/Documents/Codex	218,615	0
2 (00:40:00-00:40:33Z)	all of ~/Documents, ~/Desktop, iCloud Drive (~/Library/Mobile Documents), ~/.local/share, ~/.local/state	138,971	0

• Targets: a59c6ed74e839ca43ee704033e207865ed938a22e4738afce22ec197e91b6bcd (the original Alpaca receipt) and adb2a240727bcd0c309c0c0ae34c649387ba23a0eba3da5d99c067d25bb06ad0 (the selected rows).
• No file named for AAPL with a 2020 path or a 202008 name was found either.
• /Volumes holds only the boot volume and mounted application images, so no external data drive was in scope.

Limit: the selected-rows hash is computed over derived rows, so a whole-file search can match it only if some file stores exactly those rows. A raw multi-session file holding the same sessions would not match it. With no AAPL 2020 file on the Mac at all, the overturn route (a hash-gated reacquisition in the trading lane's next data wave) stands.

Evidence on the Mac: ~/.local/state/native-agent-stack/evidence/aapl-g4-hash-search-20261008/ (search.py, both reports, SHA256SUMS).

[seathatflowsinourveins]

Claude session native-agent-stack-5f: landing at head 5bdbd641aa3aae2507a5e771616e07dadeb62f96. The command center (wsl-architecture-design) gave its ACK at this exact head; under the user's 2026-10-04 decision it owns merges that touch the hot registry. The read-only cross-family review (GPT-6.1 Sol max, the packaged Codex SDK worker) returned ACCEPT at this exact head, with no P1 item.

Observed main 2cad4d7407e0497d4eaa51dc3c4410ef112cd928. coordination/merge_tree_landing_check.py (sha256 a11649fcf2e5844c…) exit 0:

main 2cad4d7407e0497d4eaa51dc3c4410ef112cd928 head 5bdbd641aa3aae2507a5e771616e07dadeb62f96 base ffd4b8508ae2494bf5bf7ed9df1932537dbb7ecf merged-tree d8a826cb2386ed8c845fb6b914f4be503cc2d11c merge-tree-exit 0
ok   1: clean three-way merge
ok   2: merged-vs-main paths 1, outside PR-owned 0 []
ok   3: main drift 11 paths, overlap with PR-owned inputs (registry excluded) []
ok   4: registry foreign rows equal True, order preserved True, PR-owned rows 0, unowned top-level keys differing []
ok   5: merged files[] sorted by path with no duplicates: True (10566 rows)
LANDABLE
condition 5: merged files[] sorted and unique (tree d8a826cb2386ed8c845fb6b914f4be503cc2d11c)

Required checks at this head: 7 pass . Unresolved review threads: 0.

[seathatflowsinourveins]

Claude session native-agent-stack-5f: post-merge observation. Landed as 90f9d32dbb9092b02ed6ca7f9d08e1e4807ad6ab (parent 2cad4d7407e0497d4eaa51dc3c4410ef112cd928), tree d8a826cb2386ed8c845fb6b914f4be503cc2d11c. The landed tree equals the checked merged tree and the parent is the observed main.