sparse-autoencoder-training

Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.

Quality

86%

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

85%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.

A thorough, highly actionable skill body with well-structured workflows and clean progressive disclosure into real reference files. The main weakness is conciseness: several conceptual/background sections restate knowledge Claude already has.

Suggestions

Trim the 'The Problem: Polysemanticity & Superposition' and 'What SAEs Learn' background sections to a one-line framing, since Claude already knows these concepts.

Remove or condense the 'Key Validation (Anthropic Research)' aside and the ASCII pipeline diagram; keep only what is needed to act.

Cut the 'External Resources' link lists unless they carry guidance not already in the reference files, to reduce token load.

Dimension	Reasoning	Score
Conciseness	Mostly efficient code and tables, but several sections explain concepts Claude already knows ('The Problem: Polysemanticity & Superposition', the Anthropic 70% validation aside, the ASCII diagram) that could be trimmed.	3 / 5
Actionability	Copy-paste-ready executable code across all three workflows plus concrete troubleshooting snippets covering the common training and analysis cases.	5 / 5
Workflow Clarity	Three workflows are clearly sequenced with numbered steps, explicit evaluation/validation checkpoints (metrics tables), and per-workflow checklists for error recovery.	5 / 5
Progressive Disclosure	Body is a clear overview with key examples inline and bulk detail split into real, one-level-deep reference files (README.md, api.md, tutorials.md) that are all present and clearly signaled.	5 / 5
	Total	18 / 20 Passed

Description

87%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 description with explicit what-and-when structure and a distinct, well-targeted niche. Minor gains available from adding more user-natural synonyms and a few more concrete actions.

Dimension	Reasoning	Score
Specificity	Names the domain and several concrete actions ('training and analyzing', 'decompose neural network activations into interpretable features') but stops short of the comprehensive multi-action list of the 5-anchor.	4 / 5
Completeness	Explicitly answers both what ('Provides guidance for training and analyzing Sparse Autoencoders...to decompose...into interpretable features') and when ('Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations').	5 / 5
Trigger Term Quality	Includes natural trigger phrases ('discovering interpretable features', 'analyzing superposition', 'monosemantic representations') but omits some common synonyms a user might say like 'SAE' or 'feature steering'.	4 / 5
Distinctiveness Conflict Risk	Occupies a clear niche (SAEs via SAELens for mechanistic interpretability) with distinct triggers and minimal overlap risk with other skills.	5 / 5
	Total	18 / 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/saelens/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.