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

57%

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 ./backend/cli/skills/ml-training/mamba/SKILL.md

Quality

Content

57%

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 complete, executable examples, but it suffers from repeated performance marketing claims that hurt conciseness and, more seriously, a broken progressive-disclosure structure where all three cited reference files do not exist in the bundle.

Suggestions

Fix the broken references: either rename/create references to match selective-ssm.md, mamba2-details.md, performance.md, or update the citations to point to the actual files (architecture-details.md, benchmarks.md, training-guide.md).

Deduplicate the '5× faster / no KV cache / linear scaling' claims so they appear once (e.g., in a single Performance or When-to-use section) rather than across three sections.

Add explicit validation checkpoints to the loading and benchmark workflows (e.g., assert output shape, confirm generate() returns tokens, sanity-check benchmark timing).

Dimension	Reasoning	Score
Conciseness	The body is mostly executable code and commands, but the '5× faster inference / no KV cache / linear scaling' claims are repeated across Expected results, Advantages, and Hardware requirements/Performance sections, padding the token budget without adding information.	2 / 3
Actionability	It provides fully executable, copy-paste-ready code (Mamba block, MambaConfig LM, from_pretrained loading, Mamba-2), concrete bash benchmarks, and specific troubleshooting commands with exact flags.	3 / 3
Workflow Clarity	Workflows are clearly sequenced and runnable, but none include explicit validation checkpoints or fix-retry feedback loops (e.g., verifying a model loaded or benchmark output is sane), so they stay at the steps-present-but-checkpoints-missing level.	2 / 3
Progressive Disclosure	The body signals three one-level-deep references (selective-ssm.md, mamba2-details.md, performance.md), but none of those files exist in ./references/ — the actual files are architecture-details.md, benchmarks.md, and training-guide.md — so every reference path is a dead link and navigation is broken.	1 / 3
	Total	8 / 12 Passed

Description

57%

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 about the Mamba architecture but reads as a spec sheet rather than invocation guidance: it answers 'what' thoroughly while omitting any explicit 'when to use' triggers, and leans on expert jargon over natural user phrasings.

Suggestions

Add an explicit 'Use when...' clause naming natural trigger situations (e.g., 'Use when working with Mamba/SSM models, long-context inference, or linear-complexity sequence modeling').

Lead with action verbs describing what the skill does (e.g., 'Build, load, and benchmark Mamba state-space models') instead of leading with architectural properties.

Soften jargon by pairing technical terms with natural phrasings users actually say ('long-context models', 'efficient inference').

Dimension	Reasoning	Score
Specificity	It names the domain richly (Mamba-1/2, d_state values, model sizes 130M-2.8B) but describes what the model IS ('State-space model with O(n) complexity', 'no KV cache') rather than concrete actions the skill performs, so it does not reach the multiple-specific-actions bar of 3.	2 / 3
Completeness	It clearly answers 'what' (state-space model, variants, sizes) but provides no 'Use when...' clause or equivalent explicit trigger guidance for when to invoke the skill, capping completeness at 2 per the rubric guideline.	2 / 3
Trigger Term Quality	It includes some natural relevant terms ('Mamba', 'Transformers', 'HuggingFace') but leans on technical jargon ('Selective SSM', 'd_state', 'KV cache') and lacks common user phrasings, falling short of good coverage.	2 / 3
Distinctiveness Conflict Risk	It occupies a clear, specific niche (Mamba/SSM architecture) with concrete variant and size details, making it unlikely to wrongly trigger for unrelated skills.	3 / 3
	Total	9 / 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: synthetic-sciences/openscience
Path: backend/cli/skills/ml-training/mamba/SKILL.md
Commit: dc00e07

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.