PRISM v2.10.0 released

[Ronkupper]

PRISM v2.10.0 is released — the Scope-Integrity Test (SIT): the enforcement half of the Lens Library's minimum_scope_binding:.

Additive and behavior-strengthening (MINOR). No capability is removed; default behavior is unchanged.

What landed

The gap SIT closes. A lens's minimum_scope_binding: states what coverage requires — but Probe 1 could mark a lens covered on an enumeration that looked complete while quietly scoping the question too narrowly. The cleanest pattern: a competitor scan scoped by product category rather than by audience and job, so a same-audience substitute in a different form factor (a hardware device, a manual workflow, a do-nothing default) goes unnamed — and more enumeration passes inside the wrong scope only harden the wrong boundary.

Coverage-time falsifier gate. At Probe 1, a lens cannot be marked fires-covered until its minimum_scope_binding: passes a yes/no falsifier — every clause satisfied with evidence, or any clause unmet? Generic for all 23 lenses on day one, documented as an honest rigor ladder: inline self-check (the always-on floor — the enumerating agent self-certifies, a single distribution, not triangulation, per SP-15) → fresh-context probe (minimum independence) → cross-vendor probe (full SP-15 triangulation). Levels 2–3 are operator-invokable today in a single chat; a future auto_drive only automates the dispatch.

New lens-schema field scope_integrity_probe:. An optional per-lens sharpened falsifier that overrides the generic gate. The embedded Lens Library advances to v0.13, shipping the first hand-authored probe on LL-D-008 ("Compared to what?"): name at least one cross-form-factor, same-audience-same-job substitute the set omits, or document its absence with rationale. The other 22 ride the generic gate; sharpened probes accrete from real misses.

Lint. lint_catalog_version → 4, reserving PRISM-LINT-09 (scope-integrity-probe-valid), sibling to the reserved PRISM-LINT-08.

Calibration / report-back

• Does the generic gate sharpen coverage without fatigue? The inline self-check runs on every fires-covered — confirm it catches under-scoping without turning Probe 1 into a checkbox slog.
• Does the LL-D-008 sharpened probe catch the category-vs-audience trap on a live engagement where the generic gate alone might not?
• Independent-probe rigor — when stakes warrant, does escalating to a fresh-context or cross-vendor probe surface omissions the inline floor missed?

Release: v2.10.0 · the canonical PRISM.md and the byte-identical snapshot PRISM_v2_10_0.md are on main · embedded Lens Library at v0.13.