Visual-Thinking-Lens

Classify Radial Collapse Prior (RCP) in AI-generated images, batches, prompts, or visual diagnostics. Use when the user provides an image, overlay, kernel readings, VTL/LSI values, RCA-2/VCLI context, or asks whether an output shows radial collapse, center-lock, symmetric void, density bowl, inflated cohesion, ring-conform gestures, emblematic collapse, single-force radial field, RCP_high, RCP_med, hard RCP, soft RCP, borderline RCP, or needs a first-pass Delta/Omega/O/Hold route.

Interpret, critique, explain, or configure VCLI-G (Visual Cognitive Load Index - Geometry Coupled) readings for images. Use when the user provides VCLI-G, SCI, G1/G2/G3/G4 channels, centroid wander, void topology, contour curvature, orientation entropy, structural coherence, perceptual load, earned tension, resolved clarity, collapse into noise, default simplicity, image complexity diagnostics, generative image reranking, or iterative visual load trajectories. Use the current absolute-cap VCLI-G math, not the older style-relative z-score/profile formula.

Audit reports, captions, interpretations, papers, slide text, methods notes, public explanations, or agent outputs that use Playground Analytic metrics or vocabulary. Use when checking whether claims from VCLI-G, SCI, VTL Kernel, LSI v2, RCA-2, RDC, RCP, perturbation, sequence, telemetry, model-comparison, or safe-middle analyses overstate the evidence, turn metrics into quality judgments, confuse related terms, ignore QA/comparability, treat single images as cohort evidence, or need safer claim language.

Run images through the canonical VTL Kernel Metrics extractor and return deterministic kernel vectors, r_v gradient-field package values, mask QA, hashes, and CSV/JSON outputs. Use when the user provides image paths, image folders, batches, or generated images and asks to measure VTL kernel metrics, extract delta_x, delta_y, r_v, rho_r, mu, x_p, theta, d_s, sdi, mass_fraction, gradient_floor_85, gradient_ceiling_97, tail_gap, EFA, mask_status, mask_mode, mask_sha256, or kernel_vec_sha256 using the latest notebook math rather than older PDF math.

Interpret current VTL Kernel Metrics vectors and extraction outputs as structural coordinates, not quality scores. Use when the user provides delta_x, delta_y, r_v, rho_r, mu, x_p, theta, d_s, sdi, mass_fraction, gradient_floor_85, gradient_ceiling_97, tail_gap, EFA, mask_status, mask_mode, kernel CSV/JSON output, or asks what a single image or batch occupies in VTL kernel space, how to read placement, density, cohesion, peripheral pull, orientation stability, structural thickness, spatial dispersion, or mask QA caveats.

Pressure-test image prompts, generated images, visual concepts, scenes, and creative directions through a standalone staged teardown: base, probe, yield, collapse echo, and suspended opposition. Use when a user wants to expose defaults, reveal what an image or prompt avoids, stress-test structural weakness, induce productive failure, diagnose over-resolution, or rebuild an idea through constraint, contradiction, rupture, and held tension without relying on any larger visual framework.

Diagnose whether markmaking, hatching, linework, surface texture, brushwork, engraving marks, drawing marks, or painterly strokes are structurally stratified instead of globally applied. Use when an image or prompt needs material-specific mark vocabularies, varied stroke roles, surface resistance, directional friction, edge discipline, pressure changes, or a critique of uniform texture, decorative hatching, global style passes, and marks that do not carry tonal, spatial, or material consequence.

Diagnose and build tonal hierarchy before expressive markmaking, engraving, drawing, painterly, or black-and-white image critique. Use when an image, image prompt, iteration sequence, or visual concept needs a value backbone: grayscale survival, silhouette/block-value readability, figure-field-focus priority, reserved whites, crushed blacks, midtone restraint, selective visibility, suppression zones, light-path logic, or a readiness check before asking for varied marks or surface style.

Create, critique, and revise image prompts toward coherent off-center basins: displaced structural mass, peripheral gravity, asymmetry pulse, dislocation, active void, edge-weighted composition, undercharged center, tilted viewfinder, weighted negative space, and stabilizers that prevent collapse back to centered harmony or empty minimalism. Use when a user wants spatially aware prompt repair, non-centered composition, off-center mass placement, frame-edge gravity, basin-style composition, or rule-of-thirds-like displacement with stronger structural pressure.

Lightweight orchestrator for choosing among Russell Parrish / A.rtist I.nfluencer visual reasoning skills. Use when a user asks which skill to use, gives an ambiguous image/prompt/model-analysis request, wants a visual critique workflow, or needs routing between Concert Mode, Sketcher Lens, Artist's Lens, Visual Thinking Lens, Marrowline Critique, Reverse Iterative Decomposition, Deformation Operator Playbook, Volumetric Container of Force, Foreshortening Recipe Book, Off-Center Prior Diagnostic, Radial Collapse Prior, Prompt Collapse Suite, and Structural Prompt Stabilizer. This skill selects the right skill or sequence and does not replace specialist skills.

Convert fuzzy image ideas, prompts, critiques, and creative direction into standalone visual specifications with intent, anchors, selections, invariants, transformations, constraints, negatives, viewfinder instructions, material rules, failure checks, and acceptance criteria. Use when a user wants a structured prompt, prompt repair, engine-agnostic visual spec, constraint stack, visual QA checklist, or a way to compile artistic intention into testable generation instructions without relying on any larger visual framework.

Diagnose when AI-generated images or prompt-output sequences regress to center-fill defaults: centered salience, subject blob plus passive margins, comfortable void, mid-lane containment, additive/global packing, regular cadence, symmetry, safe polish, generic clarity, and closure rescue. Use when the user wants to identify native AI visual defaults, compare default versus anti-default behavior, score default severity, name the protected attractor, or propose minimal anti-default correction moves without relying on other skills.

Revise AI image/video prompts with precise μ nudges that escape centered model defaults while preserving fidelity. Use when a prompt or generated result is too centered, too frontal, too filled, too blank, overdrifted, over-stylized, or losing identity, and the user needs a one-step correction using subject offset, luminance bias, gaze angle, footprint, negative-space ratio, symmetry caps, and fidelity locks.