mamba-architecture

State-space model with O(n) complexity vs Transformers' O(n²). 5× faster inference, million-token sequences, no KV cache. Selective SSM with hardware-aware design. Mamba-1 (d_state=16) and Mamba-2 (d_state=128, multi-head). Models 130M-2.8B on HuggingFace.

Quality

68%

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

Fix and improve this skill with Tessl

tessl review fix ./01-model-architecture/mamba/SKILL.md

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 content is highly actionable with executable code throughout, but it is held back by repeated performance claims, implicit workflow checkpoints, and a broken reference structure where linked files do not exist and the real bundle files are orphaned.

Suggestions

Fix the reference links: the Advanced-topics section points to selective-ssm.md, mamba2-details.md, and performance.md, none of which exist, while architecture-details.md, benchmarks.md, and training-guide.md are in the bundle but never referenced — align the links to the actual files.

Consolidate the duplicated '5× faster / no KV cache / linear scaling' claims that recur in Workflow 4, Advantages, Hardware requirements, and Performance into a single performance section.

Add explicit validation checkpoints to the workflows, e.g. verify the model is loaded and on CUDA before calling generate, and assert output shape/dtype after generation.

Dimension	Reasoning	Score
Conciseness	The body is mostly actionable code rather than padded prose, but the '5× faster inference', 'no KV cache', and 'linear vs quadratic' claims are restated across Workflow 4, Advantages, Hardware requirements, and Performance, so it could be tightened by consolidating them into one section.	2 / 3
Actionability	It provides extensive copy-paste-ready, executable code and commands — Mamba block usage, Mamba-2 LM config with generation, pretrained-model loading, Mamba-1 vs Mamba-2 setup, benchmark commands, and concrete troubleshooting fixes.	3 / 3
Workflow Clarity	Workflows are sequenced (install → configure → generate; load → generate) and a Common-issues section offers error-recovery hints, but there are no explicit validation checkpoints within the workflows, so checkpoints are implicit rather than enforced.	2 / 3
Progressive Disclosure	The body has good section structure and clearly-signaled one-level reference links in Advanced topics, but every referenced file (selective-ssm.md, mamba2-details.md, performance.md) is absent from the bundle while the actual provided files (architecture-details.md, benchmarks.md, training-guide.md) are never linked, so cross-file navigation is broken.	2 / 3
	Total	9 / 12 Passed

Description

72%

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 concrete and distinctive with strong trigger terms, but it reads as a spec sheet rather than an action list and omits any explicit 'when to use' guidance, capping completeness at 2.

Suggestions

Add an explicit 'Use when...' clause naming triggers, e.g. 'Use when building or loading Mamba/state-space models for long-context or efficient inference, or when comparing against Transformers.'

Reframe capabilities as concrete actions the skill performs — 'Load pretrained Mamba models (130M–2.8B), build Mamba-1/Mamba-2 language models, and benchmark generation vs Transformers' — rather than a list of architectural properties.

Dimension	Reasoning	Score
Specificity	Phrases like '5× faster inference', 'million-token sequences', 'no KV cache', and 'Mamba-1 (d_state=16) and Mamba-2 (d_state=128, multi-head)' are concrete and non-vague, but they are framed as architectural properties/specs rather than the concrete user actions (e.g. load pretrained models, build LMs, benchmark) the top anchor calls for, and the action set is not comprehensive.	2 / 3
Completeness	It clearly answers 'what' (state-space model, O(n), faster inference, model sizes) but contains no 'Use when...' clause or equivalent explicit trigger guidance, which caps completeness at 2 per the rubric guidelines.	2 / 3
Trigger Term Quality	It prominently includes the natural terms a user would actually say when needing this skill — 'Mamba', 'state-space model', 'SSM', 'Transformers', 'KV cache', 'HuggingFace' — giving good coverage of the primary trigger vocabulary.	3 / 3
Distinctiveness Conflict Risk	'Mamba', 'Selective SSM', and 'state-space model' define a clear, specific niche that is unlikely to trigger for the wrong skill or conflict with other skills.	3 / 3
	Total	10 / 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
relative_links	Relative link issues: 3 missing	Warning
referenced_paths_exist	Referenced path issues: 6 missing	Warning

	Total	13 / 16 Passed

Repository: Orchestra-Research/AI-Research-SKILLs
Path: 01-model-architecture/mamba/SKILL.md
Commit: 773a529

Reviewed: about 5 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.