Accepted real-RDP batch: 82a658a

July 18, 2026 ยท View on GitHub

Candidate 82a658a6926ddac74b010b613535c023d0b5f079 ran one fresh, fixed batch of exactly three trials from retained clean qualification base {8f43c385-0566-44d3-9e50-f646777d315b}. There were no retries.

The task was to open the Windows Run dialog through RDP, type a command that creates one trial-unique file, and verify that file through an independent prlctl exec type <path> guest-tools oracle. All three expected values matched the independently observed values exactly.

Environment and readiness

  • Substrate: real Aardwolf RDP 0.2.14 into a Parallels Windows 11 guest (Microsoft Windows [Version 10.0.22631.6199]), 1280x800 framebuffer.
  • Base source commit: db87e3ffe802a94046f0f131da6094dac9a0fbd7.
  • Qualification account: one exact active target-account RDP session; all trials qualified session 3, with Explorer in that same exact session.
  • Desktop gate: the fixed-VM Windows 11 light taskbar occupied the bottom 8%, with at least 50% of pixels at luma 161 or greater; framebuffer transition from the login baseline was at least 0.10 unless the baseline was already desktop-ready; three consecutive ready frames changed by no more than 0.02.
  • Counted desktop-readiness timeout: 75 seconds per trial.

Counted rows

TrialExpectedIndependently observedExactInput receiptLatency
1oaflow-rdp-trial-1-a7afe4a26c8135afoaflow-rdp-trial-1-a7afe4a26c8135afYesReturned without error51.845 s
2oaflow-rdp-trial-2-fae5492e869de211oaflow-rdp-trial-2-fae5492e869de211YesReturned without error10.467 s
3oaflow-rdp-trial-3-367fd897497feee3oaflow-rdp-trial-3-367fd897497feee3YesReturned without error7.477 s
MetricResult
Successes3/3
Failures0
Silent incorrect successes0
Over-halts0
Model calls0

The declared failure taxonomy was connect_or_frame_failure, input_delivery_failure, independent_oracle_mismatch, over_halt_or_timeout, and environment_restore_failure. All three counted rows were classified none.

Evidence and cleanup

The exact raw counted report remains local with SHA-256 7c31e220bfba34057b83a8910181ce54ab82bd67f5bbc527fb319ad2091b6b9b. The committed results_82a658a_20260718.sanitized.json is a deterministic derivative that retains the evidence-relevant environment, rows, counters, readiness, and cleanup booleans while replacing local machine identifiers and snapshot-ID arrays. The committed sanitized file SHA-256 is c6be771a567dfef8e924655ac165ea78c1eb32893bcf441908f4c6bc8c4c4246. It embeds the raw source-report hash, an explicit redaction manifest, and verifiable self-excluding derivative-payload SHA-256 f1c5c4b9c83f99f89a0a5039c0fcbb7dca9ed458cccd49e37bc961e645565ae2.

Final cleanup passed. It deleted only the batch-owned snapshot, restored the exact eight-ID snapshot inventory, and left the VM suspended. The current snapshot pointer was then switched without resume to the unchanged original immutable base {35dba943-a22d-473c-b1b0-44fa6326e626}. The retained clean qualification base and both footage branches remain present.

This evidence qualifies the tested RDP transport and input path for this exact task and environment. It does not qualify Citrix ICA/HDX, general desktop workflow reliability, or any untested Windows application.