Owner guide extraction and command transport source controls

This is finite harness evidence, not #46 native owner-walkthrough acceptance.

The milestone/owner-walkthrough worktree preserves the exact bytes and source spans of the two operator guides. It rejects changed admitted guide bytes, ambiguous section anchors, unterminated fences and archive overwrite. The SSH transport records the fresh-shell boundary explicitly, passes pasted secrets only through stdin, retains raw output in mode0600 files, and refuses further phases after a command failure. Actual guest boot/package/semantic joins and the complete walkthrough driver remain incomplete.

Clean source	Actual command (inside resource launcher)	Result
3139fad2d1b033c1cd9c1d331c3329d4af43ba40	env PYTHONPATH=tools/qualification python3 -B -m unittest -v test_owner_guide	8 tests, exit0
0bb47787be628d8f89fe986e831adce7f0d7cd2d	env PYTHONPATH=tools/qualification python3 -B -m unittest -v test_owner_guide test_owner_commands	13 tests, exit0

Both commands were launched with FABRIC_STORAGE_ROOT=/run/media/kmosoti/data/FabricO11y and python3 tools/resource_group.py -- from the data-drive worktree. Both outer commands exited0, confirmed service stop and removed their owned temporary scratch. The default16GiB high/20GiB maximum and no-swap limits were enforced. The respective memory peaks were18.2MiB and24.9MiB. No VM or product service ran. The literal transport failure control really exited7 and verified its captured stderr and refusal to continue; this is not a native guest fault result.

Evidence directories:

• results/owner-guide-source-01: run.log, resource.json; resource SHA256 a19f7693fe103d355f826b1b6145618a300ec2c998ef9b43cea37da70e2b6081.
• results/owner-guide-source-02: run.log, resource.json; resource SHA256 2158cd66147f51c45d65d59f568b58757acc068db7331d8046feb4a33e142ea2.

These generated source controls supplement the registered protocol; they are not independent delivery/query oracles. No product package was changed, and neither an extracted command nor an exit0 establishes the required native semantic observations. Documentation/integration checks and the genuine walkthrough remain pending. The replacement candidate required by the newly failing rustls dependency gate must be frozen before package acceptance resumes.

Accumulated phase-budget correction

The original transport assumed the protocol table imposed a strict phase order. That assumption was rejected before native acceptance: the table specifies required observations and per-phase budgets, while actual guide operations revisit phases. Prospective red at e3425f18071948684709a2c5d91191341d65d3c1 ran4 methods with4 capability/guard errors (exit1), not successful negative detections; failed evidence is retained. Separate policy1329657 and implementation316ab5226e6df68b147494d08763bf0fc437c5bd accumulate each phase budget with a hard1800second overall limit.

Exact corrective command:

python3 -B /run/media/kmosoti/data/FabricO11y/results/owner-walkthrough-driver/phase-green-driver.py owner-phase-order-green-01 1073741824 env PYTHONPATH=tools/qualification:tools python3 -B -m unittest -v test_owner_guide test_owner_commands test_owner_phase_resume test_owner_evidence

Actual21methods,zero errors/skips,exit0 at20:36:00Z; source clean before/after; outer stop and temporary cleanup confirmed. Outer receipt SHA2561913b92e1974f758b90bf048b780e863d4fe77571a91ed31ff7eca3d111e2a2f; log SHA2565750cce4a94ef67a342704091583e26ec823ffab3c2141e3b05e0919c3a9ddc8. Default16/20GiB,noSwap,1800seconds and1GiB DATA reservation remained enforced.

These source tests do not authenticate production snapshots. Subsequent source inspection found the legacy gN-M snapshot adapter disagrees with production opaque snapshots; a separately registered bound adapter and actual stopped raw-journal/Spool evidence are still required before native walkthrough grading. No native #46 acceptance is claimed.