optimizing-attention-flash

Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.

Quality

81%

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

71%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.

Highly actionable, well-sequenced content with strong code examples, but it is padded with redundant checklists and repeated commands, and half of its advanced-topic reference links point to files that are missing from the bundle.

Suggestions

Create the missing references/algorithm.md and references/advanced-features.md files (or remove the dead links) so all four Advanced topics links resolve.

Remove the per-workflow copy-paste checklists that restate the steps that immediately follow, and consolidate the repeated 'pip install flash-attn --no-build-isolation' into a single install note.

Merge the duplicate benchmarking snippets in Workflow 1 Step 3 and Workflow 2 Step 4 into one shared example to reduce token weight.

Dimension	Reasoning	Score
Conciseness	Mostly code-dense and avoids explaining concepts Claude already knows, but includes redundant padding: per-workflow checklists that restate the immediately-following steps, the install command 'pip install flash-attn --no-build-isolation' repeated three times, and near-duplicate benchmark blocks in two workflows.	3 / 5
Actionability	Provides copy-paste-ready, executable code and exact commands throughout — SDPA usage, backend forcing, profiling, accuracy comparison, flash-attn variants, and FP8 conversion — covering the common cases comprehensively.	5 / 5
Workflow Clarity	Each of the three workflows is explicitly sequenced with checklists and includes validation checkpoints (Step 3 verify speedup, Step 4 test accuracy matches baseline), plus a Common issues section for error recovery; minor gaps only.	4 / 5
Progressive Disclosure	Good overview structure with clearly signaled one-level-deep references, but two of the four referenced files (references/algorithm.md and references/advanced-features.md) do not exist in the bundle, breaking navigation for half the Advanced topics links.	3 / 5
	Total	15 / 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.

A strong, third-person description that concretely states what it does and when to use it, with good natural-language triggers and a distinct niche. Only minor improvement possible in trigger-term synonyms.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions and capabilities — 'Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction' plus 'Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention' — giving comprehensive coverage of what the skill does and which backends it covers.	5 / 5
Completeness	Explicitly answers both what ('Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction') and when ('Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference') with concrete trigger phrases.	5 / 5
Trigger Term Quality	Strong natural triggers users would say — 'training/running transformers with long sequences', 'encountering GPU memory issues with attention', 'need faster inference' — but misses a few common variations like 'OOM', 'attention is slow', or 'FlashAttention'.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche — Flash Attention optimization for long sequences and GPU memory pressure — with distinct triggers unlikely to fire for unrelated skills.	5 / 5
	Total	19 / 20 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 2 missing	Warning
referenced_paths_exist	Referenced path issues: 4 missing	Warning

	Total	13 / 16 Passed

Repository: OpenLAIR/dr-claw
Path: skills/optimization/flash-attention/SKILL.md
Commit: 4b35756

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.