fix(trading): make paper qualification proofs optional and mark receipts

[seathatflowsinourveins]

Scope

• What this PR changes, in one or two sentences: runner.py paper starts without archived qualification proofs, and supplying either proof still requests the complete promotion check. The paper receipt's preflight block records promotion_gate (validated or not_supplied), as the mover runner's receipt does.
• Base commit: 03a986ed637349403658f23da694c1833527819d. The branch is one commit on it.
• Lane: lane:trading.
• Owned paths touched:
  • blueprints/us-equities/adaptive-paper/runner.py, README-paper-exploration.md and source-hashes.json;
  • tests/test_adaptive_paper_runner.py;
  • blueprints/us-equities/data/README.md, whose stale mandatory-gate text this corrects;
  • the matching manifests/evidence.json registrations.

SOTA sources

• native-agent-stack mover_runner.py:703-704 at the base above. Line 703 is the optional proof-pair selection adopted here, and line 704 is the promotion_gate receipt marker adopted here. The mover's recover command (command_recover, mover_runner.py:885) sets no marker, and neither does this runner's recover.
• runner.py:661 and runner.py:2310 at the base: qualification dispatch, and the prior unconditional CLI selection. The account, margin, session, inventory and reconciliation checks stay in their established paths.
• Registration follows the docs/lanes.md hot-file protocol (scripts/host_receipts.py register_file). source-hashes.json was recomputed from the current bytes of the files it lists, after a byte-identical serialization round trip.

Evidence-class table

Claim	Evidence class	Command / receipt
Paper without proofs reaches post-gate execution, and its preflight summary records promotion_gate: not_supplied	local_integration (fixture credentials, hermetic broker observation)	MainCommandGateWiring; its marker assertions failed first, see below
A blocked account without proofs still refuses with account_not_ready, and its written receipt records not_supplied	local_integration	same
A valid complete proof pair passes and records validated; either proof alone and invalid complete proofs refuse	local_integration	same
recover adds no marker; preflight is unchanged	local_integration	same
Registered hashes match the head bytes	local_integration	source-hash manifest tests, evidence_manifest.py --check, validate.py
Measured broker-paper acceptance is NOT_RUN	source_review	No account was selected and no broker runtime was started

Local commands run

The engine environment was built with the README-native.md install command, uv pip sync --python "$ENGINE_PYTHON" --require-hashes blueprints/us-equities/adaptive-paper/requirements-linux-x86_64-py312.lock. It has Python 3.12.4, exchange-calendars 4.13.2 and nautilus_trader 2.0.0rc5.

$ "$ENGINE_PYTHON" -B -m unittest tests.test_adaptive_paper_runner.MainCommandGateWiring   # new tests, previous head's runner.py (blob 4951f603)
exit 1: Ran 10 tests, FAILED (failures=3): None != 'not_supplied' (both no-proof tests), None != 'validated' (valid-proof test)
$ "$ENGINE_PYTHON" -B -m unittest tests.test_adaptive_paper_runner.MainCommandGateWiring   # with this change
exit 0: Ran 10 tests, OK
$ "$ENGINE_PYTHON" -B -m unittest tests.test_adaptive_paper_runner
exit 0: Ran 187 tests, OK
$ "$ENGINE_PYTHON" -B -m unittest tests.test_adaptive_paper_credential_source tests.test_adaptive_paper_fees tests.test_alpaca_admission_regressions tests.test_adaptive_paper_mover tests.test_adaptive_paper_mover_native
exit 0: Ran 219 tests, OK   # the other modules that drive runner.main()
$ python3 -B -m unittest tests.test_adaptive_paper_source_hashes_manifest tests.test_trading_gates_bindings
exit 0: Ran 20 tests, OK
$ python3 scripts/evidence_manifest.py --check
exit 0: {"files": 11423, "status": "passed"}
$ python3 scripts/validate.py
exit 0: {"components": 70, "hashed_files": 11423, "profiles": 4, "receipts": 239, "status": "passed"}
$ gitleaks git --log-opts="03a986ed637349403658f23da694c1833527819d..HEAD"   # merge-base with origin/main
exit 0: 1 commits scanned, no leaks found
$ python3 scripts/validate.py --scan-history 03a986ed637349403658f23da694c1833527819d HEAD
exit 0: {"matching_locations": 0, "status": "passed"}

Decision record

blueprints/us-equities/adaptive-paper/README-paper-exploration.md records the exploration direction, the pinned mechanism and the test evidence. It states preset and leverage-rung coverage as the direction, not current behavior. The existing config-hash continuity check still refuses same-account preset or rung rotation. No runner or configuration alias for the exploratory paper runs has been verified, and the inspected T22 snapshot supplies no production paper dispatcher.

Alternative considered: setting the marker on every paper/recover path. That would stamp not_supplied on recover receipts for trials that a validated paper run admitted. It was rejected because the mover's recover sets no marker. If recover receipts later need to restate the admitting run's qualification, that would be a change to carry the original marker forward, not to recompute it.

Host evidence

No host-evidence files or platform statuses change.

Checklist

• New/changed GitHub Actions are pinned to a full commit SHA with a
  version comment (no floating tags). No workflow or action changes.
• New/changed workflows declare top-level permissions: {} and grant each job
  only what it needs (contents: read, or an explicitly justified addition). No workflow changes.
• No secrets are printed, logged or committed; no new required secret was
  added without a documented owner.
• No new paid hosting, subscription or billing surface was introduced.
• Peer-owned untracked files and worktrees were preserved (not deleted,
  moved or overwritten).

🤖 Generated with Claude Code