Skill audit report

sparse-autoencoder-training Audit report.

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.

EXPERIMENTAL · REVIEWNeeds reviewGenerated Oct 12, 2026Heuristic metadata audit

Audit

Trust

Quality

Security

100

Maintain

Install

OpenAgentSkill Trust Score

Strong shortlist

OpenAgentSkill Trust Score

The Trust Score helps an agent decide whether a skill is safe enough to shortlist before installation.

GitHub adoption

PASS

100

13K GitHub stars

Stars/forks activity

PASS

13K stars, 948 forks; issue activity unavailable in current metadata

Recent maintenance

PASS

100

1d since push

License clarity

PASS

MIT

README/SKILL.md completeness

PASS

Metadata includes enough usage and workflow context

Dependency/runtime risk

FAIL

command execution surface, credential or environment access

Install availability

PASS

npx skills add Orchestra-Research/AI-Research-SKILLs --skill sparse-autoencoder-training

Install command safety

PASS

standard package or runtime install path

Permission surface

FAIL

secrets or environment access, shell or command execution

Repository evidence

PASS

https://github.com/Orchestra-Research/AI-Research-SKILLs/tree/main/04-mechanistic-interpretability/saelens

Review status

WARN

50.9

AI review approval is missing

Agent Proven outcomes

INFO

No agent outcome data yet

Checks

Install and adoption review

8 Passed · 11 Needs review

Install path

PASS

npx skills add Orchestra-Research/AI-Research-SKILLs --skill sparse-autoencoder-training

Repository

PASS

https://github.com/Orchestra-Research/AI-Research-SKILLs/tree/main/04-mechanistic-interpretability/saelens

License

PASS

MIT

Maintenance

100

PASS

1d since push

AI review

CHECK

Review approval is missing

README/SKILL.md completeness

PASS

Usable description available

Dependency risk

FIX

command execution surface, credential or environment access

Install command safety

PASS

standard package or runtime install path

Permission surface

FIX

secrets or environment access, shell or command execution

Stars/forks activity

PASS

13K stars, 948 forks; issue activity unavailable in current metadata

Adoption

PASS

13K GitHub stars

Warnings

Dependency or permission surface needs review
Permission surface may require sandboxing
AI review approval is missing
Quality score needs review
Permission surface needs review: secrets or environment access, shell or command execution
Dependency/runtime risk: command execution surface, credential or environment access
Permission surface: secrets or environment access, shell or command execution
Review status: AI review approval is missing

Method

This report combines public metadata, AI review output, repository freshness, install readiness, OpenAgentSkill events, quality scoring, trust checks, and the agent safety gate. It is not a full source-code security review.

Install path

Review the report before installing into production agents.

$ npx skills add Orchestra-Research/AI-Research-SKILLs --skill sparse-autoencoder-training

Agent-readable metadata

Machine decision packet

Embedded JSON exposes suited tasks, suited agents, install command, trust score, safety gate, alternatives, and do-not-use conditions for agent review.

Install policy: review
Best agents: Codex + Claude Code
Alternatives: 3

Open agent JSON

Signals

Audit score: 84/100
Quality: 87/100
Trust: 78/100
Install safety: 92/100
Permission surface: 22/100
Maintenance: 1d since push
Events: 2 events
GitHub stars: 13K
Last push: Oct 11, 2026

Agent safety v2

44/100 · Avoid automatic install

Experimentalreview

Sparse or mixed signals. Useful for discovery, but not for autonomous installation.

Test manually in an isolated workspace and compare against safer alternatives.

Metadata combines secrets access with shell or command execution
High-risk permission hints: Shell or command execution, Secrets or environment access
44/100 agent safety score

Shell or command execution

high

Skill metadata references terminal, CLI, shell, subprocess, or command execution workflows.

Network access

medium

Skill likely fetches remote pages, APIs, repositories, or external services.

Filesystem access

medium

Skill may read or write project files, documents, generated artifacts, or local workspace state.

Secrets or environment access

high

Skill metadata references credentials, tokens, environment variables, or secret-bearing workflows.

High-risk permission hints: Shell or command execution, Secrets or environment access
Dependency or permission surface needs review

Back to skill Alternatives Audit badge SVG

Compare nearby options

Related skills to audit next

weather

Current weather and forecasts with web_fetch, falling back to wttr.in curl for locations, rain, temperature, travel planning.

391K Stars · Audit report

mcporter

List, configure, authenticate, call, and inspect MCP servers/tools with mcporter over HTTP or stdio.

391K Stars · Audit report

release-openspec

Use this skill when releasing OpenSpec: audit merged work and changeset coverage, decide whether a catch-up changeset PR is needed, prepare or resume the Changesets Version Packages PR, cut a beta or stable release, verify publishing, and polish GitHub release notes. Also use when asked whether an open release PR is complete, what the next release step is, or to continue a release paused for human approval.

71K Stars · Audit report