Conciseness

The body is dense and earns its tokens — design choices with numbers ("local_kernel=7", "gate initialized at 0.3", "pad_size=32... 12.5% extension"), comparison tables, and no tutorials on basics Claude already knows. A few spots could be trimmed (the 'Why reflection padding works' bullets partially restate the section intro, and "This analysis reveals the fundamental spectral limit" is a filler transition), which keeps it at 4 rather than the lean anchor at 5.

Actionability

The frequency_band_errors function is fully executable and the FNOBlock class is near-complete with concrete hyperparameters, but ShockTubeFNO1d is deliberately abbreviated ("# ... standard FNO layers ..." and literal `...` in __init__) and SpectralConv1d is used without definition — minor gaps versus the copy-paste-ready anchor at 5.

Workflow Clarity

This is a design-selection skill rather than a linear procedure, and the "Architecture Selection for Different Shock Problems" table plus the viscosity/resolution table give an unambiguous problem → architecture → hyperparameter mapping. Not a batch/destructive skill, so no validation cap applies; it stays at 4 rather than 5 because there is no explicit training/evaluation sequence or checkpoints for applying the techniques end-to-end.

Progressive Disclosure

No bundle files exist and the body is a well-sectioned, self-contained ~140-line overview with clear headers — appropriate structure for a single-file skill. The 'Future Directions (from literature)' table is borderline padding that could live in a references file, and at this length a split (e.g., full architecture code into a reference file) would be reasonable, which are the minor gaps that hold it at 4.

Specificity

The description lists four concrete capabilities — "local-global spectral design (ShockFNO), reflection padding for non-periodic BCs, resolution scaling for shock width, and frequency-band error diagnostics" — which is comprehensive coverage of the skill's scope, matching the top anchor rather than the 'several specific actions; minor gaps' anchor at 4.

Completeness

Both required parts are explicit: 'what' via "Architectures and techniques for neural operators on discontinuous PDE solutions... Covers [four techniques]" and 'when' via "Use for low-viscosity Burgers, compressible Euler, Riemann problems, or any PDE where standard FNO produces Gibbs oscillations" — the exact structure of the anchor-5 example.

Trigger Term Quality

Natural trigger terms a domain user would actually say are well covered: "low-viscosity Burgers, compressible Euler, Riemann problems", "shocks, contact discontinuities", "standard FNO produces Gibbs oscillations". These include problem names and failure-mode phrasings that directly match how a user would voice the need, with no obvious synonyms missing for this niche.

Distinctiveness Conflict Risk

This is a clear niche (shock-capturing neural operators) with distinct triggers — specific equations (Burgers, Euler), specific model family (FNO), and a specific artifact (Gibbs oscillations). It would not plausibly fire for a neighboring skill; overlap risk is minimal.