hqq-quantization

Half-Quadratic Quantization for LLMs without calibration data. Use when quantizing models to 4/3/2-bit precision without needing calibration datasets, for fast quantization workflows, or when deploying with vLLM or HuggingFace Transformers.

Quality

84%

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

76%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 highly actionable and well-structured with executable examples and real reference files, but workflows lack explicit validation checkpoints before save/push-to-Hub operations, and a few inline sections duplicate reference content. Adding a verify-before-commit step and trimming repeated setup would lift the weakest dimensions.

Suggestions

Add an explicit validation/verification checkpoint in the quantize-and-save and push-to-Hub workflows (e.g., run a generation sanity check before save_pretrained/push_to_hub) to satisfy the workflow_clarity cap for batch/destructive operations.

Trim the duplicated installation and basic-config blocks between Quick start and Core concepts, or consolidate them into one canonical snippet to improve conciseness.

Move the detailed backend list and mixed-precision per-layer table into references/advanced-usage.md, keeping only a short pointer in SKILL.md, to strengthen progressive disclosure.

Dimension	Reasoning	Score
Conciseness	The body is largely efficient with code-first guidance and minimal concept explanation, but the Quick start repeats concepts (installation, basic config) that reappear in later sections, offering modest trim potential.	4 / 5
Actionability	Copy-paste ready, executable code examples cover the common cases (basic quantization, HuggingFace integration, vLLM, PEFT/LoRA, mixed precision) with specific imports and parameters.	5 / 5
Workflow Clarity	Workflow examples are numbered and sequenced, but they omit validation/verification checkpoints before destructive or batch operations; the only 'verify quality' step is optional generation output, and saving/pushing to Hub has no validation gate.	3 / 5
Progressive Disclosure	Good structure with a clear overview and two real, one-level-deep references (advanced-usage.md, troubleshooting.md) that exist in the bundle, though some advanced content (backends, mixed precision) could arguably move into the reference rather than staying inline.	4 / 5
	Total	16 / 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 strong, concise, and explicit, clearly stating both what HQQ does and when to use it with concrete trigger phrases and named frameworks. Minor improvement would come from adding common synonyms like 'model compression' or 'low-bit' to broaden trigger coverage.

Dimension	Reasoning	Score
Specificity	Names concrete actions (quantizing models to 4/3/2-bit precision) plus specific deployment targets (vLLM, HuggingFace Transformers) and workflows, giving comprehensive coverage.	5 / 5
Completeness	Explicitly answers 'what' (Half-Quadratic Quantization for LLMs without calibration data) and 'when' with an explicit 'Use when ...' clause listing concrete trigger scenarios.	5 / 5
Trigger Term Quality	Includes natural user phrases like 'quantizing models', 'calibration datasets', 'fast quantization', and framework names (vLLM, HuggingFace), but is missing common synonyms such as 'compress model', 'reduce memory', or 'low-bit'.	4 / 5
Distinctiveness Conflict Risk	The calibration-free, fast, specific-bit-precision niche with named frameworks is highly distinct from other quantization skills and unlikely to trigger for the wrong skill.	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/hqq/SKILL.md
Commit: fcd2e11

Reviewed: 2 days 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.