long-context

Extend context windows of transformer models using RoPE, YaRN, ALiBi, and position interpolation techniques. Use when processing long documents (32k-128k+ tokens), extending pre-trained models beyond original context limits, or implementing efficient positional encodings. Covers rotary embeddings, attention biases, interpolation methods, and extrapolation strategies for LLMs.

Quality

82%

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

Quality

Content

65%

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

The body is highly actionable with executable code throughout, but it is verbose and somewhat monolithic: concept explanations and code overlap, and detailed sections duplicate content available in the referenced files rather than deferring to them. Workflows lack explicit validation checkpoints.

Suggestions

Trim the 'Core Concepts' math/advantages block and the repeated scaling-factor examples, since the Quick Start and Implementation Patterns already cover them, to improve conciseness.

Add an explicit validation checkpoint to the fine-tuning workflow (e.g., measure perplexity on held-out long documents before increasing the scaling factor).

Move the full 'Fine-tuning for Long Context' and 'Production Deployment' detail into references/fine_tuning.md, keeping SKILL.md an overview with the existing See Also links to strengthen progressive disclosure.

Dimension	Reasoning	Score
Conciseness	Mostly actionable and efficient, but the 'Core Concepts' section re-explains RoPE/ALiBi math and advantages that overlap the Quick Start and Implementation Patterns code, and several scaling-factor examples repeat, so it could be tightened despite avoiding pure fluff.	2 / 3
Actionability	Provides multiple copy-paste-ready executable blocks (RotaryEmbedding class, ALiBi bias functions, rope_scaling configs, Trainer fine-tuning args, vLLM deployment), with specific values rather than pseudocode.	3 / 3
Workflow Clarity	A method-choice -> scaling -> fine-tuning -> deployment sequence is implied in Best Practices, but there are no explicit validation checkpoints (e.g., verify perplexity on long docs before scaling further), only implicit ✅/❌ guidance.	2 / 3
Progressive Disclosure	Three real one-level-deep references are signaled in 'See Also', but the SKILL.md body is a heavy, fairly monolithic inline guide whose fine-tuning and deployment content overlaps references/fine_tuning.md rather than being split into it.	2 / 3
	Total	9 / 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.

The description is strong: it states concrete capabilities, gives explicit 'Use when...' triggers covering both what and when, uses natural domain terms, and occupies a distinct niche. It avoids fluff and over-claims.

Dimension	Reasoning	Score
Specificity	Names multiple concrete actions ('Extend context windows', 'processing long documents', 'extending pre-trained models', 'implementing efficient positional encodings') alongside specific techniques (RoPE, YaRN, ALiBi, position interpolation), matching the 'lists multiple specific concrete actions' anchor.	3 / 3
Completeness	Explicitly answers both what ('Extend context windows... using RoPE, YaRN, ALiBi, and position interpolation') and when ('Use when processing long documents..., extending pre-trained models..., or implementing efficient positional encodings'), matching the clear-both anchor.	3 / 3
Trigger Term Quality	Natural trigger phrases a user would say appear throughout ('long documents (32k-128k+ tokens)', 'extending pre-trained models beyond original context limits', 'context windows'), plus the domain's standard technique names, giving good coverage.	3 / 3
Distinctiveness Conflict Risk	It carves a clear niche (transformer long-context extension) with distinct, specific triggers unlikely to fire for unrelated skills; voice is appropriately third person with no over-claims.	3 / 3
	Total	12 / 12 Passed

Validation

81%

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 — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (537 lines); consider splitting into references/ and linking	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 1 missing	Warning

	Total	13 / 16 Passed

Repository: Orchestra-Research/AI-Research-SKILLs
Path: 19-emerging-techniques/long-context/SKILL.md
Commit: 773a529

Reviewed: about 11 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.