Anti-Slop Labs

GitHub Copilot hands-on labs · VS Code + GitHub Copilot app

Less AI slop. More proof.

Two hands-on labs. Practice the built-in /commands and Spec Kit, then the skills that stop guessed fixes, tests that prove nothing and unverified "done" claims. Every exercise starts from broken code and ends with an independent check.

Set up the labLab 1Lab 2

Runs locally in VS Code or the GitHub Copilot desktop app. No npm install, API key or cloud resource.

Illustrative walkthrough: Make tests red. Fix the cause. Prove green. Useful tests fail on the starter and pass on the repair; the lab checks both.
Illustrative GIF · Before · After · Transcript

The two labs

Do them in order or pick one. Both share the same local environment and work in either client.

Set up your lab

  1. Install the prerequisites

    Node.js 22.20+ and Git, plus VS Code signed in to GitHub Copilot or the GitHub Copilot app. Lab 1's Spec Kit part also needs uv.

  2. Get the lab

    The ZIP holds the broken starters, the independent checks and ten pinned public skill folders. Extract it and open a terminal in the anti-slop-lab folder.

    Download the lab (ZIP) Public skills only (ZIP) GitHub repository

    Or clone it:

    git clone https://github.com/AkashAi7/anti-ai-slop.git
cd anti-ai-slop/lab

  3. Check your machine and create a workspace

    Each track gets a fresh Git repository under work/ with the broken starter and a baseline commit.

    npm.cmd run doctor -- builtins
npm.cmd run setup -- builtins

    Tracks: builtins, spec-kit, spec-kit-skills (Lab 1) and skills (Lab 2). On macOS or Linux, use npm.

  4. Open the workspace in your client

    ClientOpen the workspaceThen
    VS Codecode -n ./work/builtins, or File > Open FolderStart a Local chat session and type /
    GitHub Copilot appAdd the work/builtins folder as a local repositoryStart an Interactive session and type /

Full setup guide: every track, the baseline check and troubleshooting

Load the anti-slop skills

Lab 2: anti-slop skills bundles 10 public skill folders from five pinned repositories (29 SHA-256-checked files). Sources: obra/superpowers @ 2ce8aa1 (MIT); mattpocock/skills @ 4ae2317 (MIT); addyosmani/agent-skills @ 8bfcc1f (MIT); blader/humanizer @ 0ef9fe5 (MIT); anthropics/skills @ ca45fc4 (Apache-2.0). They are public skills, not Copilot built-ins.

SkillSlop it prevents
/grill-meTurn a vague request into confirmed decisions before any code is written.
/systematic-debuggingReplace speculative fixes with a reproduced root cause.
/test-driven-developmentMake a regression fail for the right reason before repairing code.
/verification-before-completionReplace a confident completion claim with fresh terminal evidence.
/receiving-code-reviewVerify each review comment against the code and brief instead of agreeing with all of them.
/code-simplificationRemove unnecessary complexity from working code without changing behavior.
/humanizerStrip AI writing tells from the prose that ships with your change.
/source-driven-developmentGround framework-specific code in cited official documentation, not memory.
/frontend-designAvoid templated, generic-looking UI by committing to a deliberate design direction.
/grillingProvides the structured interview used by /grill-me.

  1. Bundled with the lab (fastest)

    Copies the pinned skill folders into work/skills/.github/skills and checks all 29 files against SHA-256 hashes. No network needed.

    npm.cmd run setup -- skills
npm.cmd run skills

  2. Your own project

    Download the public skills ZIP and copy its .github folder into your repository root. Or fetch the latest upstream with the skills CLI, then review what it installed:

    npx skills add mattpocock/skills --skill grill-me --skill grilling -a github-copilot

  3. Confirm the client loaded them

    VS Code: /skills opens Configure Skills. GitHub Copilot app: /skills reload, then /skills. Every skill must point at the workspace copy.

Lab 2, Part B: all loading options and checks

Running it as a workshop