Agent Skills: Exercise Builder

|

UncategorizedID: willoscar/research-units-pipeline-skills/exercise-builder

Install this agent skill to your local

pnpm dlx add-skill https://github.com/WILLOSCAR/research-units-pipeline-skills/tree/HEAD/.codex/skills/exercise-builder

Skill Files

Browse the full folder contents for exercise-builder.

Download Skill

Loading file tree…

.codex/skills/exercise-builder/SKILL.md

Skill Metadata

Name
exercise-builder
Description
|

Exercise Builder

Goal: attach at least one verifiable exercise to every module so the tutorial has a teaching loop.

Inputs

  • outline/module_plan.yml

Outputs

  • Updated outline/module_plan.yml

Exercise schema (recommended)

For each module, add an exercises list. Each exercise should contain:

  • prompt
  • expected_output
  • verification_steps (a checklist)

Workflow

  1. Read outline/module_plan.yml and enumerate modules.
  2. For each module, design ≥1 exercise that directly verifies the module objectives.
  3. Ensure every exercise has an expected output and a verification checklist.
  4. Update outline/module_plan.yml in place.

Definition of Done

  • [ ] Every module in outline/module_plan.yml has ≥1 exercise.
  • [ ] Every exercise includes expected_output + verification_steps.

Troubleshooting

Issue: exercises are open-ended with no verification

Fix:

  • Convert them into “do X → observe Y → verify Z” with concrete artifacts.

Issue: exercises drift from the running example

Fix:

  • Re-anchor each exercise to the module’s running_example_steps so the tutorial stays coherent.