autoregressive-neural-pde-solver

Training patterns for autoregressive neural PDE solvers (FNO, DeepONet, CNO). Covers rollout training, noise injection for stability, multi-component loss functions (H1, frequency-sensitive, boundary-aware), per-channel normalization for coupled multi-variable systems, and the PDEBench nRMSE metric. Use when training any neural operator that predicts time-dependent PDE solutions.

Quality

90%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

80%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A token-efficient, highly actionable reference packed with executable patterns and prescriptive tables. Its main gaps are the absence of an explicit sequenced workflow with validation checkpoints and a lack of progressive disclosure via separate reference files.

Suggestions

Add a short numbered end-to-end training workflow with explicit validation checkpoints (e.g. train → check val nRMSE → verify both nRMSE formulas beat baseline → report) to lift workflow_clarity.

Move the longer, self-contained code blocks (loss implementations, the dual nRMSE metric) into reference files under references/ and link to them from SKILL.md to introduce genuine one-level-deep progressive disclosure.

Make the validation/feedback loop explicit in the nRMSE section — state the pass condition and what to do if a baseline is not beaten.

Dimension	Reasoning	Score
Conciseness	The body is dense and assumes competence — code, tight commentary, and tables — without explaining concepts Claude already knows (no 'what is a PDE' or library introductions), so most tokens earn their place.	3 / 3
Actionability	It provides copy-paste-ready, executable code with concrete shapes, weights, and noise levels (e.g. NOISE_STD values, h1_loss, frequency_loss) plus a recommended-combinations table, matching the fully-executable anchor.	3 / 3
Workflow Clarity	Sections are logically ordered (principle → data → noise → loss → normalization → metric → split → pitfalls) but there is no explicit step sequence with validation checkpoints; validation is only implicit ('verify you beat baselines under both'), so it sits at the level-2 anchor rather than 3.	2 / 3
Progressive Disclosure	The document is well-sectioned but monolithic — all content is inline in SKILL.md with no bundle files or one-level-deep references, and at ~190 lines it exceeds the under-50-line exception, so the level-2 anchor (good structure, content that could be split is inline) fits better than 3.	2 / 3
	Total	10 / 12 Passed

Description

100%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A precise, third-person description that names concrete capabilities and provides an explicit usage trigger with natural domain terminology. It cleanly satisfies the what/when pair and is highly distinct.

Dimension	Reasoning	Score
Specificity	The description lists multiple concrete capabilities — 'rollout training, noise injection for stability, multi-component loss functions (H1, frequency-sensitive, boundary-aware), per-channel normalization... PDEBench nRMSE metric' — matching the anchor for enumerating several specific actions.	3 / 3
Completeness	It answers both what ('Training patterns... Covers rollout training, noise injection...') and when via an explicit 'Use when...' clause, satisfying the top anchor and avoiding the missing-trigger cap at 2.	3 / 3
Trigger Term Quality	It surfaces the natural domain vocabulary a practitioner would actually say — 'FNO, DeepONet, CNO', 'neural operator', 'PDE solutions' — plus the explicit 'Use when training any neural operator that predicts time-dependent PDE solutions' trigger.	3 / 3
Distinctiveness Conflict Risk	The niche — autoregressive neural PDE solvers with named operator families — is narrow and clearly distinct, with triggers unlikely to fire for unrelated skills.	3 / 3
	Total	12 / 12 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/physics/autoregressive-neural-pde-solver/SKILL.md
Commit: 52845c3

Reviewed: about 10 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.