C1f: record the first real-world case; untrack two unread Termux binaries

[b-macker]

Summary

Documentation and housekeeping from the repo-sentinel dogfood round 5 (live Gemini, NAAb 80c8cd9). There are no code changes.

• docs/open-investigations.md, row C1f: records the first real-world case of S5's startup-frozen entropy baseline.

  • One arm ran four times, two with the clean fixture and two with the adversarial one. In all four the reviewer holds 1.0000 through turn 5, pays S5 0.15 every turn from turn 6, and ends at 0.556667 and OUTPUT_INADMISSIBLE at turn 8, identical to six decimals whatever the model wrote.
  • The first firing is set by vocab_contraction_window (6), which is why changing the baseline window from 3 to 5 didn't move it.
  • Every analyzed row in all eight runs reconciled from telemetry (CDD: reconcile coherence from telemetry; analyze the response after a failed call #263).
  • The row's "no shipped example is known to be affected" is marked superseded.

• run-all-tests.sh: corrects the test_signal_contract.sh skip note. It said the gate fails all four criteria, but C1/C2 now pass and only C3/C4 fail. The gate stays out of the sweep.

• Two binaries untracked, with .gitignore entries:

  • docs/book/.../Vigilant/proxy/gateway_vessel
  • docs/book/.../Vigilant/scanner/shield_vessel

  They're Android aarch64 builds, about 9.4 MB, which AUDIT-REPORT.md already flagged. Nothing reads them: synthesizer.naab builds into Vigilant/bin/ from the gateway.go / shield.rs sources committed beside them, and only when that binary is missing.

The default change this evidence argues for is in a separate PR, so it can be decided on its own.

Test Plan

• The only executable change is a comment in run-all-tests.sh.
• The note's new claim was re-checked by running test_signal_contract.sh on the current build: C1 passes, and an earlier run showed C1/C2 pass and C3/C4 fail.
• The two untracked binaries have no references anywhere in the tree except AUDIT-REPORT.md, which was checked by grep.

🤖 Generated with Claude Code

https://claude.ai/code/session_01ELUfjXZvx8kzXo1UJjrAhC

---

Generated by Claude Code

[github-actions]

NAAb Governance Report

Metric	Count
Files checked	16
Passed	16
Failed	0

All governance checks passed!

_{Generated by NAAb Governance Engine v4.0}