neuron-explorer
Steering-first neuron explorer for Qwen3.5-4B: push every MLP neuron, keep every rollout, browse the results
翻译准备中
尚未检测到更新。GitHub 发布版本后会显示在这里。
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证Measured Agent Skills for Claude Code and other agents. Every Skill ships its provenance, and the measured ones ship their eval.
翻译准备中
取自仓库 README
正在加载完整 README。 以下为已保存的摘录。
Twenty-three Agent Skills, and a loop that works out which of them are actually worth keeping.
--list shows everything without installing. A Skill is just a folder with a SKILL.md at its root, so cloning and copying works too:
What if every time your AI agent failed, it automatically got better?
Four of these Skills form a loop that runs on your own machine, over your own session logs, and only keeps a Skill that measurably improves your own work:查看完整 README ↗Edge gives your AI agent the right expert skill for the task at hand. It searches 100,000+ security-screened public skills, picks the best fit, and loads a pinned version for that task. Works in Claude Code, Codex, Cursor and the Claude app.
Using Claude Code or Codex? Add skill discovery:
npx skills add getedgehq/skills --skill edge-find-skills
Set up Edge: paste this into Claude Code, Codex or Cursor:
Set up Edge for me: fetch https://getedge.cc/SKILL.md?ref=gh-skills with curl, read it, and follow it
After setup, your agent can load any skill in this repository automatically when a task needs it. Details on getedge.cc.
Explore Edge's benchmark evidence: published results, study protocols, task manifests and reanalysis scripts, alongside credited external research. Browse the interactive benchmarks for outcomes and method details.
These are examples returned by production find_skill on 2026-09-29; they are third-party catalog results, not packages maintained in this repository. Results can change.
| Your task | Skill Edge returned first |
|---|---|
| Write a CFO outbound email sequence | onewave-ai/claude-skills@cold-email-sequence-generator |
| Create an investor pitch deck | manojbajaj95/claude-gtm-plugin@pitch-deck-creation |
| Analyze a sales CSV | claude-office-skills/skills@data-analysis |
| Write a competitor positioning brief | phuryn/pm-skills@competitor-analysis |
Edge scans packages, mirrors available third-party scanner reports, and loads pinned revisions. A scan reports findings within its stated scope; it is not a safety guarantee. Browse the skill evidence hub or individual pages for product launch video, autonomous research, and workplan.
Browse on Edge · How evidence works · Apache 2.0 license
This repository contains 161 top-level skill bundles, including compatibility aliases. It is the versioned source for Edge-maintained skills. The public database holds package provenance and recorded evaluations where they exist. Browse the package catalog and evidence notes for installation and benchmark details.
Your agent publishes personal deck links and tells you who opened, what they actively read, and when another device appeared. Original system and design by Falco Schneider; Edge ports it into a Node + SQLite skill. See tracked-deck-links for setup and provenance.
Original tracking system and design by Falco Schneider. Apache-2.0 under the repository default licence; approved 2026-10-07.
Turn your location data into an interactive globe with search, filters, glowing clusters, entity drilldown and mobile sheets. data-globe includes the reusable template and normalization tools. Explore the live reference: Rocketmap. Apache-2.0.
Use Rocketlist’s current startup hiring data inside your agent: role, location, stage, salary and remote filters, with direct employer links. The public dataset and Python helper need no account or API key.
| Need | Skill |
|---|---|
| Find current startup jobs and hiring companies | rocketlist-job-search |
This bundle is licensed under Rocketlist’s MIT license; the hiring dataset has its own upstream data terms.