Probe debt tracker: unprobed rules and clauses (round-5 queue)

[F-e-u-e-r]

Standing tracker for the house covenant's probe debt: every behavioral rule
or clause shipped with an unprobed marker owes a bare-vs-ruled probe (or a
recorded demotion/decline). This issue is the re-read surface that keeps the
debt visible — per skill-authoring §3, a debt parked only in a plan or
summary nothing re-opens is an invalidation failure.

Live view (canonical — do not hand-maintain a list here)

The in-body markers in the skill files are the ground truth:

grep -rn 'unprobed' skills/*/SKILL.md

Snapshot at a148180 (occurrence counts per file; includes Provenance prose —
the in-body markers are the canonical subset):

cross-model-review 4 | delegation-and-review 34 | domain-evidence 1
ground-truth-gates 14 | operational-rigor 28 | security-architect 4
skill-authoring 40 | skill-vetting 1

Recent additions: +11 unprobed rules/clauses from the #85–#89 combined merge
(#90); #91 (pre-registration ordering clause) will add one more if merged.

Closure semantics

A marker leaves the queue only by one of:

1. Probe run — bare/ruled arms per skill-authoring §7 (environment-
   relative baseline, audience tier, symmetric leakage controls), and the
   marker updated in place with the result. Precedent: skill-authoring: repair the derived-file rule's trigger (it fails both arms), record first probe results for the 2026-07-25 four #85 recorded first
   probe results for the four 2026-07-25 rules, including one repair and one
   flagged non-discrimination.
2. Demotion/decline — a recorded judgment that the line comes out or the
   probe is deliberately declined, with the reason in Provenance.

Probes are owner-run on the private successor suite (see the README
covenant); results land here as marker diffs, not as suite artifacts.

This issue stays open as long as any in-body unprobed marker exists; it is
a queue, not a milestone.

[F-e-u-e-r]

The staged security-enhancement library (PR #106) ships 6 in-body unprobed OPS markers (OPS-2/6/8/9/10/11). Since this tracker's ground truth is grep unprobed skills/*/SKILL.md (installed skills) and that library is staged, not installed, its deferred items — the six markers plus a one-sentence MANIFEST covenant sync (SEC-9-1) — are tracked separately in #105 to keep this tracker's installed-skills scope clean. Cross-linked for visibility.

[F-e-u-e-r]

+3 in-body unprobed markers land with #108 (ae0b108): ground-truth-gates item 10 (a green suite names the artifact it exercised), the ground-truth-gates detector-positives bullet, and skill-authoring §7's probe-scenario rule ("attested round as shape"). All three carry Provenance entries naming the round-5 queue.

Fresh occurrence-count snapshot at 82bb983 (occurrence counts incl. Provenance prose — the in-body markers remain the canonical subset per this issue's grep):

cross-model-review 4 | delegation-and-review 34 | domain-evidence 1
ground-truth-gates 24 | operational-rigor 31 | security-architect 4
skill-authoring 42 | skill-vetting 1