Mindket

VibeSkills

Newversion 1 · updated
Star3821

Overview

Picks the right skills from a large installed pool and breaks each task into checkable steps for your coding agent. In its own benchmark, mean reward rose 21.12 percentage points and token use fell 29.6%.

VibeSkills is an open-source project on GitHub in the Developer category, listed by Mindket. It is free. This is version 1, last updated on . Install it from GitHub as its README says. Get Mindket’s setup notes on your Mindket account, as a zip.

About

What it is

VibeSkills is a general-purpose skill that routes your coding agent to the right local skills and plans how they run together. When many skills are installed, an agent can pick the wrong ones or repeat work. This skill sorts out the task first, then chooses only the few skills that matter.

Who it suits

It suits people who run coding agents with many skills installed and want the agent to choose and order them sensibly.

Who uses it and for what
IndustryWho uses itHow they use itExample
SoftwareDevelopers who use coding agentsInstall it so the agent selects relevant skills from a big poolAn agent that picks a few matching skills instead of calling many
AI engineeringTeams that benchmark agentsCompare the same tasks with and without itA paired comparison of an agent's task results
AutomationTeams that chain several toolsSplit a complex task into subtasks that can each be verifiedA multi-step job with a check before delivery

If you only use a handful of skills, routing among them adds little, so you can skip it.

Example

You give your agent a complex task with a clear deliverable. VibeSkills first restates the goal and what must be delivered. It splits the work into subtasks it can verify, picks a few fitting skills from the pool and runs them in the order their dependencies require.

For newcomers and experts

In plain terms, it makes the agent think before it acts. It confirms what you want and what counts as done, then plans small steps and checks the work before handing it over. A good first try is one multi-step task, so you can see how the agent plans it.

The project's benchmark is an adapted SkillsBench in which every task sees all 195 specialized skills at once, not just its own. The log analysis says the gain comes from selecting fewer, better-matched skills, not from calling more of them. The authors also report less rereading of context and less repeated work than a native agent. The reward analysis covers 82 paired tasks.

  1. Clarify the goal and delivery needs
  2. Split the task into verifiable subtasks
  3. Pick a few relevant skills
  4. Run them in dependency order
  5. Check the work before delivery

Good to know

  • License: Apache-2.0
  • The results come from the project's own run of version 4.1.0 with DeepSeekV4Flash-VE and OpenHands, so treat them as author-reported
  • Token use fell from 491.122M to 345.756M and tool calls fell from 9,954 to 6,664
  • Of the paired tasks, 23 came out better with it, 55 matched and 4 came out worse
  • The repository publishes machine-readable results and a public study so others can reproduce the comparison
SkillsBench results reported by the project
Mean reward without VibeSkills50.3 percent
Mean reward with VibeSkills71.4 percent
Full-score rate without VibeSkills47.6 percent
Full-score rate with VibeSkills69.5 percent

How to use it

Install it from GitHub as its README says. Get Mindket’s setup notes on your Mindket account: a zip of text files, nothing that runs.

README and files

README
# VibeSkills

Picks the right skills from a large installed pool and breaks each task into checkable steps for your coding agent. In its own benchmark, mean reward rose 21.12 percentage points and token use fell 29.6%.

Made by foryourhealth111-pixel on GitHub. These notes are Mindket's, written from its README; the project's own files and license are in its repository.

## About

What it is: VibeSkills is a general-purpose skill that routes your coding agent to the right local skills and plans how they run together. When many skills are installed, an agent can pick the wrong ones or repeat work. This skill sorts out the task first, then chooses only the few skills that matter.

Who it suits: It suits people who run coding agents with many skills installed and want the agent to choose and order them sensibly.

```table
Title: Who uses it and for what
Industry | Who uses it | How they use it | Example
Software | Developers who use coding agents | Install it so the agent selects relevant skills from a big pool | An agent that picks a few matching skills instead of calling many
AI engineering | Teams that benchmark agents | Compare the same tasks with and without it | A paired comparison of an agent's task results
Automation | Teams that chain several tools | Split a complex task into subtasks that can each be verified | A multi-step job with a check before delivery
```

If you only use a handful of skills, routing among them adds little, so you can skip it.

Example: You give your agent a complex task with a clear deliverable. VibeSkills first restates the goal and what must be delivered. It splits the work into subtasks it can verify, picks a few fitting skills from the pool and runs them in the order their dependencies require.

For newcomers and experts: In plain terms, it makes the agent think before it acts. It confirms what you want and what counts as done, then plans small steps and checks the work before handing it over. A good first try is one multi-step task, so you can see how the agent plans it.

The project's benchmark is an adapted SkillsBench in which every task sees all 195 specialized skills at once, not just its own. The log analysis says the gain comes from selecting fewer, better-matched skills, not from calling more of them. The authors also report less rereading of context and less repeated work than a native agent. The reward analysis covers 82 paired tasks.

```flow
Clarify the goal and delivery needs
Split the task into verifiable subtasks
Pick a few relevant skills
Run them in dependency order
Check the work before delivery
```

Good to know:

- License: Apache-2.0
- The results come from the project's own run of version 4.1.0 with DeepSeekV4Flash-VE and OpenHands, so treat them as author-reported
- Token use fell from 491.122M to 345.756M and tool calls fell from 9,954 to 6,664
- Of the paired tasks, 23 came out better with it, 55 matched and 4 came out worse
- The repository publishes machine-readable results and a public study so others can reproduce the comparison

```chart
Title: SkillsBench results reported by the project
Mean reward without VibeSkills: 50.3 percent
Mean reward with VibeSkills: 71.4 percent
Full-score rate without VibeSkills: 47.6 percent
Full-score rate with VibeSkills: 69.5 percent
```

## On GitHub, 16 November 2026

- Stars: 3,821
- Forks: 345
- Watching: 121
- Open issues and pull requests: 45
- Language: Python
- License: Apache-2.0
- Topics: agent-framework, agent-skills, agentic-coding, ai-agents, ai-scientist, ai-skills, ai-workflow, automation, claude-code, codex, context-engineering, developer-tools

## Links

- Code: https://github.com/foryourhealth111-pixel/vibe-skills
- Releases: https://github.com/foryourhealth111-pixel/vibe-skills/releases
- Issues: https://github.com/foryourhealth111-pixel/vibe-skills/issues
- Star history: https://www.star-history.com/#foryourhealth111-pixel/vibe-skills&Date

## Get the code

```
git clone https://github.com/foryourhealth111-pixel/vibe-skills.git
```

Then follow the setup in its README. Nothing in this kit runs by itself.
Files
README.md  4 KB

1 file, 4 KB, downloaded as a zip.

What’s new

  • Version 1 · 2026-11-16

Compare with similar

Listed GitHub projects in the same category that share the most topics with this one.

ProductStarsActivityLanguageLicenseLast change
VibeSkills This page3,821ActivePythonApache-2.0
Cheap AgentShares agentic-coding · developer-tools · ai-agents310ActiveTypeScriptMIT
ArchifyShares developer-tools · codex · agent-skills94,756ActiveJavaScriptApache-2.0
RufloShares ai-skills · codex · ai-agents81,187——MIT—
code-review-skillShares developer-tools · codex · agent-skills441ActivePythonMIT
UI UX Pro MaxShares ai-skills · codex · claude-code146,944ActivePythonMIT
Numbers as Mindket last read them, from 2026-11-07 to 2026-11-16. Activity is read from each project’s last change on GitHub: Active within 30 days, Quiet within 180, Stale after that.

Reviews

No ratings yet

Write a review

No reviews yet.

What people say

Posts and comments that name this project on Hacker News, Bluesky, X, Reddit, Facebook and YouTube, found by a weekly search and on Trendshift. Mindket doesn’t write or check them. Open one to read it where it was posted.

Searched

More from Mindket

See all from Mindket

all-mpnet-base-v2

sentence-transformers · Kit

Turns sentences and paragraphs into 768-number vectors so you can search, cluster and compare text by meaning. It suits developers and data teams working in En…

New

multilingual-e5-base

intfloat · Kit

Turns text in about 100 languages into vectors so you can compare meaning across them, for search and matching. Made for developers who build multilingual retr…

New

sprite-gen

aldegad · Skill

Turns one base drawing of a character into game-ready sprites, either an atlas with a manifest or transparent motion loops, for game developers working in Code…

New4097

Similar projects

See all in Developer

Archify

tt-a1i · Skill

Generates interactive diagrams and explanatory pages from codebases or textual plans. It allows AI agents to visualize workflows, architecture, and complex log…

New94.8k

Ponytail

dietrichgebert · Skill

Guides AI coding agents to write minimal code by prioritizing existing repository components and standard libraries. It adds auditing, code review, and trackin…

New187k

Graphify

graphify-labs · Skill

Converts codebases, documentation, and technical files into a searchable knowledge graph for AI coding assistants. It uses local AST parsing to map relationshi…

New145k