andreaswasita/copilot-agents-dojo/template

Template Skill

Two to three sentences stating what this skill does AND what it deliberately does NOT do. The "does not" half is as important as the "does" half — it prevents over-triggering on adjacent tasks.

When to Use

Specific triggers, keywords, and contexts that should activate the skill. Be slightly pushy — agents undertrigger more often than they overtrigger.

• Trigger condition 1 (concrete situation, not abstract)
• Trigger condition 2
• Trigger condition 3 (include negative triggers if helpful: "but NOT when…")

Prerequisites

What must be true before the skill can run.

• Tool / MCP server availability
• Required env vars or config keys
• Repo state (e.g. clean working tree, on a feature branch)

If a prerequisite is missing, the skill must fail loudly — never degrade silently.

How to Run

One-screen overview of the canonical happy path. If the skill ships a script, show the command:

scripts/example.sh --flag value

Quick Reference

The agent should be able to act from this table alone for routine cases.

Procedure

Step-by-step detailed workflow. This is where the depth lives.

Step 1: Assess

What the agent checks before acting.

Step 2: Act

The actual work. Reference real Copilot tools in backticks (view, edit, grep, glob, powershell, web_fetch, task). Do NOT instruct the agent to call bare shell utilities (cat, sed, find) that have a Copilot wrapper.

Step 3: Confirm

Intermediate verification before moving on.

Pitfalls

Imperative "DO NOT" register of failure modes this skill prevents.

• DO NOT [common mistake 1] — [why it bites].
• DO NOT [common mistake 2] — [why it bites].
• DO NOT [common mistake 3] — [why it bites].

Verification

How the agent (or a reviewer) proves the skill worked.

• [ ] Test or check 1 (with concrete command)
• [ ] Test or check 2
• [ ] scripts/verify.sh passes