mochammadshenna/go-adk-q/how-to

How-to: Add an agent skill

Goal: Create a new skill SKILL.md file that the /skills command surfaces in the TUI, and optionally wire it to an ADK agent or toolset.

What is a skill?

In this repo, a skill is a Markdown specification file (SKILL.md) that defines a reusable agent behaviour: its purpose, trigger conditions, inputs, outputs, and step-by-step process.

Skills live in skills/ and are organised by domain:

skills/
├── agents/
│   ├── analyzer/SKILL.md
│   ├── comparator/SKILL.md
│   └── grader/SKILL.md
├── collaboration/
│   ├── code-review-consensus/SKILL.md
│   └── ...
├── design/
├── documentation/
├── engineering/
└── workflow/

The ADK SkillToolset (tool/skilltoolset/) loads these files and exposes them as tools that the main agent can invoke.

Step 1 — Create the skill directory and file

mkdir -p skills/engineering/my-skill

Create skills/engineering/my-skill/SKILL.md:

# my-skill
 
## Purpose
 
One sentence: what this skill does and when to use it.
 
## Trigger conditions
 
-User asks to do X
-Agent output contains Y pattern
-After completing step Z
 
## Inputs
 
| Field | Type | Description |
|---|---|---|
| `context` | string | Background information |
| `target` | string | The thing to act on |
 
## Process
 
1.**Analyse** the input for X, Y, Z.
2.**Draft** a response following the template below.
3.**Review** against the checklist.
4.**Output** the final result.
 
## Output format
 
\`\`\`
## Summary
...
 
## Details
...
\`\`\`
 
## Checklist
 
-[ ] Addressed all inputs
-[ ] Output matches format
-[ ] No hallucinated facts

Step 2 — Register the skill (optional)

If you want the skill to be callable as an ADK tool, add it to the localskilltoolset initialisation in cmd/tui/main.go:

import localskilltoolset "go-adk-q/tool/skilltoolset"
 
// Inside buildRunner():
skillset, err := localskilltoolset.New(localskilltoolset.Config{
    SkillsDir: "skills",
})

The SkillToolset discovers all SKILL.md files under SkillsDir automatically — no explicit registration per skill is needed.

Step 3 — Verify in the TUI

go run ./cmd/tui chat

Type /skills — your new skill should appear in the list.

Tips

Keep SKILL.md agent-legible. LLMs read these files. Use clear headings, short sentences, and explicit enumerated steps.

Checklist = done criteria. The checklist at the end is evaluated by the agent to determine whether the skill completed successfully.

Benchmark mode. The Analyzer skill supports a benchmark mode where it compares outputs across two runs. If your skill produces structured output, consider adding a similar mode.

Related

• Reference: Skills API
• Explanation: Architecture — skill toolset