1. Conduid
  2. Gaming
  3. Tars Skills
MCP server · Gaming

Tars Skills

TARS Skill Hub - Community skill registry for TARS personal assistant

Unclaimed gaming
34Low

Scored 6 months ago · breakdown

About Tars Skills

Tars Skills is an MCP server in the Gaming category: tARS Skill Hub - Community skill registry for TARS personal assistant.

Install

From source
git clone https://github.com/devlikebear/tars-skills

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

✦

Ask AI about Tars Skills

Powered by Claude · Grounded in docs

I know everything about Tars Skills. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

  • ✓README presentThe repository ships documentation.

5 checks have not been run yet.

README

TARS Skill Hub

Public skill registry for TARS.

Usage

# Search skills
tars skill search project

# Install a skill
tars skill install project-start

# List installed skills
tars skill list

# Update all skills
tars skill update

# Search trusted MCP packages
tars mcp search

# Install a hosted MCP server
tars mcp install safe-time

# Review and install a domain pack
tars pack install github-maintainer-pack

Skill Format

Each skill lives in skills/<name>/SKILL.md with YAML frontmatter:

---
name: my-skill
description: "What this skill does"
user-invocable: true
recommended_tools:
  - tool_name
---

# My Skill

Skill instructions in Markdown...

Registry Format

registry.json indexes all available skills, plugins, and trusted MCP packages:

{
  "version": 3,
  "skills": [
    {
      "name": "skill-name",
      "description": "...",
      "version": "0.6.0",
      "author": "username",
      "tags": ["tag1", "tag2"],
      "path": "skills/skill-name",
      "user_invocable": true,
      "quality": {
        "score": 85,
        "last_updated": "2026-05-01",
        "tests_passing": true,
        "required_tools": ["bash"],
        "permissions": ["filesystem"],
        "companion_cli": true,
        "install_count": 120
      }
    }
  ],
  "mcp_servers": [
    {
      "name": "safe-time",
      "path": "mcp-servers/safe-time",
      "manifest": "tars.mcp.json",
      "files": [
        {
          "path": "tars.mcp.json",
          "sha256": "<sha256>"
        }
      ]
    }
  ],
  "packs": [
    {
      "name": "github-maintainer-pack",
      "description": "GitHub maintainer workflow bundle for log triage, issue filing, and PR/worktree operations.",
      "version": "0.1.0",
      "author": "devlikebear",
      "tags": ["github", "maintenance", "dogfooding"],
      "skills": ["github-ops", "log-watcher", "log-anomaly-detect"],
      "plugins": [],
      "mcp_servers": [],
      "quality": {
        "score": 86,
        "last_updated": "2026-05-01",
        "tests_passing": true,
        "required_tools": ["bash", "git", "gh"],
        "permissions": ["filesystem", "github", "shell", "docker"],
        "companion_cli": true
      }
    }
  ]
}

Quality Metadata

Every installable registry entry may include a quality object. TARS renders this metadata in the Extensions Hub before install:

  • score: required when quality is present; integer from 0 to 100.
  • last_updated: ISO date (YYYY-MM-DD) for the packaged entry metadata or files.
  • tests_passing: whether the latest maintainer smoke/unit test pass is known.
  • required_tools: local executables or CLIs the package expects.
  • permissions: user-impacting capabilities such as filesystem, github, network, browser, or mcp.
  • companion_cli: whether the skill ships or depends on a companion CLI/script.
  • install_count: optional observed install count when available.

Domain Packs

packs are reviewable bundles of existing skills, plugins, and MCP packages. TARS prints the install plan before applying it, then installs each member through the same sandbox checks as individual package installs.

Contributing

  1. Create skills/<your-skill>/SKILL.md
  2. Or create mcp-servers/<your-server>/tars.mcp.json plus hosted runtime files
  3. Or create a packs entry that references existing packages
  4. Add an entry to registry.json with quality metadata
  5. Open a PR

OpenClaw Compatibility

TARS can also install skills from ClawHub via adapter. Use the --source openclaw flag:

tars skill search --source openclaw <query>
tars skill install --source openclaw <package-name>

License

MIT

README copy captured when this server was indexed; it may lag the repository. The original is authoritative.

Questions

About Tars Skills

How do I install Tars Skills?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is Tars Skills safe to use with an AI agent?

Its trust score is 34 out of 100 (low). It passes 1 of the 1 static security checks run so far. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Tars Skills still maintained?

Conduid hasn't recorded a commit date for this repository yet. Check the repository directly for recent activity.