model-pruning

Reduce LLM size and accelerate inference using pruning techniques like Wanda and SparseGPT. Use when compressing models without retraining, achieving 50% sparsity with minimal accuracy loss, or enabling faster inference on hardware accelerators. Covers unstructured pruning, structured pruning, N:M sparsity, magnitude pruning, and one-shot methods.

Quality

69%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/ml-training/model-pruning/SKILL.md

Quality

Content

46%Weight 40%Scale 1-5

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

The content is actionable with concrete code but suffers from verbosity and a failure to use the provided reference bundle for progressive disclosure. Destructive pruning workflows also lack validation feedback loops.

Suggestions

Add an explicit validation checkpoint in the production pipeline (e.g., verify achieved sparsity and run a quick perplexity check before saving) so destructive/batch operations get a validate-fix-retry loop.

Reference the existing references/wanda.md from the body and move the duplicated strategy snippets and the performance comparison table into one-level-deep reference files to reduce inline bulk.

Trim redundant explanations of basic concepts (structured vs unstructured, magnitude pruning rationale) and consolidate the three overlapping pruning-strategy examples into one canonical workflow.

Dimension	Reasoning	Score
Conciseness	At ~487 lines the body restates concepts Claude already knows (structured vs unstructured basics, magnitude pruning comments) and repeats the same ideas across three near-duplicate strategy snippets, making it noticeably verbose.	2 / 5
Actionability	Provides concrete, mostly executable code (wanda_prune, nm_prune, production pipeline) covering common cases, with only minor gaps from undefined helper stubs (load_calibration_data, train_step, prune_model).	4 / 5
Workflow Clarity	A numbered production pipeline exists, but destructive/batch pruning operations lack explicit validation checkpoints or validate-fix-retry feedback loops, which caps workflow clarity at 3 per the rubric.	3 / 5
Progressive Disclosure	A bundled references/wanda.md exists but is never linked from the body, and large blocks (multiple code variants, full performance tables) are inlined rather than split into one-level-deep reference files.	2 / 5
	Total	11 / 20 Passed

Description

92%Weight 40%Scale 1-5

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

The description is specific, complete, and clearly distinguishes the skill with both a "what" and a concrete "Use when" clause. Trigger terms are natural and largely comprehensive, with only minor synonym gaps.

Dimension	Reasoning	Score
Specificity	Lists multiple specific concrete actions and techniques — "Reduce LLM size and accelerate inference" plus "unstructured pruning, structured pruning, N:M sparsity, magnitude pruning, and one-shot methods" — giving comprehensive coverage.	5 / 5
Completeness	Explicitly answers both what ("Reduce LLM size and accelerate inference using pruning techniques...") and when ("Use when compressing models without retraining, achieving 50% sparsity...") with concrete trigger phrases.	5 / 5
Trigger Term Quality	Strong natural terms a user would say ("compressing models without retraining", "50% sparsity", "faster inference", "Wanda and SparseGPT"), with only a few synonyms/file-extension-style terms missing.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche — LLM pruning via named methods (Wanda, SparseGPT) — with distinct triggers and minimal overlap with adjacent compression/quantization skills.	5 / 5
	Total	19 / 20 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/ml-training/model-pruning/SKILL.md
Commit: fcd2e11

Reviewed: 1 day 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.