Week 45, updated 2 Nov 2026

AI use cases Agent skills & MCP Give AI agents new abilities with MCP and skills

Give AI agents new abilities with MCP and skills

Extend an AI agent with MCP servers (tool connectors) and Agent Skills (instruction packs). Prompts, workflow and the best-scored options with safety labels.

Tools compared
4
Open source
3of 4
Reviewed
16 Oct

For: Developers

Best-scored tools for this jobscore, safety
  1. 1Agent SkillsAgent skill53No issues found
  2. 2MCP reference serversMCP server48No issues found
  3. 3awesome-mcp-serversDirectory47No dependencies scanned
  4. 4wshobson/agentsAgent skill42No issues found
What the score and shapes mean

Score = public maintenance, adoption and docs signals, as a percentile of 52 tools. Nobody here tested these hands-on. Shapes are the separate safety label: No issues found, No dependencies scanned, Caution.

What it does for you

People already using Claude Code, Codex, Cursor or a similar agent who want it to do more.

  • Connect an agent to services through the Model Context Protocol
  • Add reusable skills for tasks you repeat
  • Browse community lists of MCP servers to find a connector
  • Install plugin bundles for several coding agents at once
  • Write your own skill for an internal process
More about this job

MCP servers let an agent call external tools such as GitHub, a browser or a docs index. Skills are folders of instructions an agent loads when a task matches. Together they turn a general chat agent into one that follows your workflows.

Workflow in 5 steps

  1. Decide the one job you want the agent to do better.
  2. Pick a connector or skill from a reviewed source and read its README and requested permissions.
  3. Install it with the minimum scope and test on a throwaway project.
  4. Check the safety label and last-push date on this site before adding anything.
  5. Review what the agent actually did, then widen access only if needed.
Full steps
  1. Decide the one job you want the agent to do better.
  2. Pick a connector or skill from a reviewed source and read its README and requested permissions.
  3. Install it with the minimum scope and test on a throwaway project.
  4. Check the safety label and last-push date on this site before adding anything.
  5. Review what the agent actually did, then widen access only if needed.

Prompts to copy

Starting points we wrote, not tested results.

List the tools and skills you currently have access to and what each is for. Which ones would you use to review a pull request?

2 more prompts

I want you to follow this release checklist every time I say "cut a release". Draft a skill file for it with a name, a description and steps.

Help me choose an MCP server for browsing documentation. Compare two options on what they can read, whether they can write, and what credentials they need.

Watch out for

  • Text an agent reads from the web, issues or documents can contain hidden instructions (prompt injection).
  • Keep API keys, tokens and customer data out of prompts and out of files the agent can read.
  • Check each tool's licence before commercial use, and check the licence of any code or text the AI reproduces.
  • Every MCP server you add is code you run with your credentials.
  • Licence to check before commercial use: Agent Skills (Apache-2.0 + proprietary).
Why these warnings?
  • Text an agent reads from the web, issues or documents can contain hidden instructions (prompt injection). Do not give an agent that reads untrusted content write access to anything important.
  • Keep API keys, tokens and customer data out of prompts and out of files the agent can read. Use a separate, low-privilege key for any agent.
  • Check each tool's licence before commercial use, and check the licence of any code or text the AI reproduces.
  • Every MCP server you add is code you run with your credentials. Prefer maintained, widely used servers and remove ones you stop using.
  • awesome-mcp-servers: code and instruction-file checks found nothing, but there is no runtime lockfile or manifest, so dependencies were not checked

Recommended tools

From our ranked data, highest score first. A high score does not mean a clean scan.

Related jobs

Sources for tool facts

anthropics/skills · modelcontextprotocol/servers · punkpeye/awesome-mcp-servers · wshobson/agents · Model Context Protocol. Data fetched 2 Nov 2026. Scores and labels are automated, relative and not a hands-on test.

Get the board every Monday

One email a week: the new top 10, the biggest score changes and every safety label that changed.

Free. Double opt-in: you confirm by email first, unsubscribe in one click.