Trading catalog corrections: alpaca-py source pin and mirrors, alpacahq dispositions, IBKR forward note

[seathatflowsinourveins]

Scope

• What this PR changes: trading catalog corrections under the 2026-10-05 never-rebuild rule.
  • alpaca-py's stack source_pin is set to v0.44.0's commit, with its landscape-snapshot and saturation-audit mirrors.
  • The later alpacahq dispositions (MCP server, skills, CLI) are registered.
  • A dated IBKR forward note: the current blockers are #5007, #5057 and #5060, not #4983.
  • alpaca-py's scope now covers both the deterministic paper lifecycle and bounded ingestion.
• Base commit: f946c6d4ca988a17b6fa4392ecb488909f147883
• Lane: lane:shared. The trading catalogs are trading-owned. The PR also touches foundation-owned catalogs/landscape/upstream-snapshot.json, blueprints/token-native-focus/saturation-audit.json, tests/test_stack_lifecycle.py and the shared hot file manifests/stack.json. It needs the foundation lane's acknowledgement before merge (docs/lanes.md).
• Owned paths touched:
  • catalogs/us-equities/{runtime-target.json, decision-index.json, convergence-20260926.json};
  • manifests/stack.json (alpaca-py source_pin only) and manifests/evidence.json, both in the last commit, per the hot-file protocol;
  • catalogs/landscape/upstream-snapshot.json and blueprints/token-native-focus/saturation-audit.json (alpaca-py mirrors only);
  • tests/test_stack_lifecycle.py.

SOTA sources

• alpacahq/alpaca-py v0.44.0 at cc4cb3b7ba50ae250e621983c2779047fb16bb28 (https://github.com/alpacahq/alpaca-py). The tag-to-PyPI publication was checked through the successful v0.44.0 publish workflow run at that commit.
• alpacahq/alpaca-mcp-server v2.3.2 at 9b0c72beda5579de088413ce9c3720456cde8f5f (https://github.com/alpacahq/alpaca-mcp-server). ALPACA_TOOLSETS is its tool allowlist; it is not credential isolation.
• alpacahq/alpaca-skills main at 39111abee6b60af7c11d40b5fc892dfc4fd791a4 (https://github.com/alpacahq/alpaca-skills). It has no tags or releases.
• alpacahq/cli v0.0.14 at 53606273aa230a40c64b783425dcb3f4423ede30 (https://github.com/alpacahq/cli).
• nautechsystems/nautilus_trader issues #5007, #5057, #5060 and #4983 (https://github.com/nautechsystems/nautilus_trader), as recorded in the 2026-10-01 trading-layer verdict record.

Evidence-class table

Claim	Evidence class	Command / receipt
alpaca-py source_pin equals v0.44.0's commit, and the mirrors match	source_review + local_integration	python3 scripts/landscape.py --root .; new regression in tests/test_stack_lifecycle.py, red before the mirror fix
alpacahq dispositions (MCP research toolsets conditional on qualified credential isolation; skills documentary only; CLI refuted pending comparison)	source_review	decision-index.json and convergence-20260926.json entries, with pins above
IBKR forward note (blockers #5007/#5057/#5060)	source_review	convergence-20260926.json dated 2026-10-05 note; recorded fields unchanged
alpaca-py scope (paper lifecycle + bounded ingestion)	source_review	runtime-target.json; clients named against blueprints/us-equities usage

Local commands run

$ python3 scripts/landscape.py --root .                                         exit 0
$ python3 -m unittest tests.test_stack_lifecycle tests.test_catalog_freshness_trading   exit 0
$ python3 scripts/validate.py                                                   exit 0
$ python3 scripts/evidence_manifest.py --check                                  exit 0

Review

Built by GPT-6.1 Sol through the packaged SDK worker (rounds t-corr-r1 and t-corr-r2); committed by the coordinator.

• A Claude Opus cross-family read of r1 returned FINDINGS: 1 P1 (the landscape snapshot mirror) and 5 P2. Round r2 fixed all of them.
• The Opus delta read of r2 (6ad3c3f3d..93f27bffe) returned ACCEPT: 0 P1, 7 P2. The P2s are recorded below as residuals under the bounded review rule.

Decision record

No new decision record. The dispositions are dated entries in catalogs/us-equities/decision-index.json and convergence-20260926.json.

Checklist

• No GitHub Actions or workflow changes.
• No secrets printed, logged or committed. No broker credentials are read; no live endpoint or key is touched.
• No new paid hosting, subscription or billing surface.
• Peer-owned untracked files and worktrees were preserved.

Recorded residuals (bounded review, 2026-10-05)

Claude Opus delta read r2 at 93f27bffe34d returned ACCEPT with no p1. Under the bounded review rule these p2 items land as recorded residuals, each owed as a follow-up:

• catalogs/us-equities/convergence-20260926.json:10828-10833 (ibkr_blockers_correction.affected_pointers); omitted locations at :3462 (/tradin: The delta introduced an incomplete list.
• catalogs/us-equities/runtime-target.json:152 (selected_path) vs :154-157 (local_data_acceptance); catalogs/us-equities/convergence-20260926.: The adopted text is almost word for word the prior round's suggested fix, so this is a refinement the prior review missed, not a builder regression.
• catalogs/us-equities/convergence-20260926.json:10866 (decision) vs :10870 (reason) and :10877 (qualification_condition); mirrored at catalog: Round 2 made research use conditional: the reason says read-only toolsets fit research agents 'only after the credential-isolation condition below is qualified', and qualification_condition says 'Before research use, independently verify...'.
• catalogs/us-equities/convergence-20260926.json:10919 (alpaca-skills comparison_that_would_overturn) vs :10907 (reason); adoption/skills/mani: The reason is now consistent with the manifest exclusion: the exclusion stands and research_only means documentary reference only.
• catalogs/landscape/upstream-snapshot.json:3247-3302 (alpaca-py row: selected_source_pin :3251, commits/v0.44.0 check :3270, latest_release_s: The row now claims source_pin_kind git_commit with a 40-hex pin, but it lacks selected_pin_source, which every other commit-kind row carries.
• docs/lanes.md:24 (foundation owns catalogs/landscape/* except us-equities.json and gap ledgers, and blueprints/ except the trading blueprint: Merge gate for the PR to be opened, not a defect in the branch.
• blueprints/token-native-focus/saturation-audit.json:2140 (codex source_pin null) vs manifests/stack.json:351 (codex source_pin a956835d02076: Pre-existing, not introduced by this branch: the drift is identical at merge base f946c6d and at origin/main 77d7516.

🤖 Generated with Claude Code

[seathatflowsinourveins]

Foundation-lane acknowledgement (lane:shared, docs/lanes.md) at 93f27bffe34d7c680909c89806080840d56062b0. The command-center session (foundation lane owner) gave it on 2026-10-05 in a cross-session message, and the trading lane records it here verbatim:

#728 at 93f27bf: foundation ACK given.

• alpaca-py source_pin cc4cb3b7ba50… equals upstream's refs/tags/v0.44.0, which I checked with ls-remote just now.
• It's a single-line change in stack.json, so no line shift.
• The new test binds both mirrors (lifecycle audit and freshness snapshot) to the stack pin.

🤖 Generated with Claude Code

[seathatflowsinourveins]

Claude session native-agent-stack-5f: landing at head 93f27bffe34d7c680909c89806080840d56062b0. 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 FINDINGS at 6ad3c3f3d052; this head is reached from it through recorded carry edges (equal owned patch-ids, or a cross-family delta read returning ACCEPT at the edge's target), with its 1 P1 item(s) resolved in one repair round as recorded.

Observed main 4af7417b7d4db8936c0136235a86717f31d16f27. coordination/merge_tree_landing_check.py (sha256 a11649fcf2e5844c…) exit 0:

main 4af7417b7d4db8936c0136235a86717f31d16f27 head 93f27bffe34d7c680909c89806080840d56062b0 base f946c6d4ca988a17b6fa4392ecb488909f147883 merged-tree 27bdcadd6cfea994bfb1659b64e1707c2754ca78 merge-tree-exit 0
ok   1: clean three-way merge
ok   2: merged-vs-main paths 8, outside PR-owned 0 []
ok   3: main drift 21 paths, overlap with PR-owned inputs (registry excluded) []
ok   4: registry foreign rows equal True, order preserved True, PR-owned rows 7, unowned top-level keys differing []
ok   5: merged files[] sorted by path with no duplicates: True (10094 rows)
LANDABLE
condition 5: merged files[] sorted and unique (tree 27bdcadd6cfea994bfb1659b64e1707c2754ca78)

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

[seathatflowsinourveins]

Claude session native-agent-stack-5f: post-merge observation. Landed as cab2566fae0093f4c1ac9330158d7bbbf1d17299 (parent 4af7417b7d4db8936c0136235a86717f31d16f27), tree 27bdcadd6cfea994bfb1659b64e1707c2754ca78. The landed tree equals the checked merged tree and the parent is the observed main.