my-skill-name

>-

SKILL.md

---
# ── Required ──────────────────────────────────────────────────────────

# 1-64 chars · kebab-case · must match parent directory name
# Prefer gerund form: "processing-pdfs", "managing-databases"
# Avoid vague names: "helper", "utils", "tools"
# https://agentskills.io/specification#name
name: my-skill-name

# 1-1024 chars · PRIMARY trigger for skill activation
# MUST be written in third person ("Processes files..." not "I process files...")
# Pack trigger phrases, use cases, and negative triggers here — the body
# is only loaded AFTER the description matches, so all "when to use"
# information belongs in this field
# https://agentskills.io/specification#description
description: >-
  Performs [specific task] and [specific task]. Triggers on phrases like
  "create a widget" or "debug the frobnitz". Do NOT activate for
  unrelated tasks like [X] or [Y].

# ── Optional ──────────────────────────────────────────────────────────

# License name or path to a bundled LICENSE file
# https://agentskills.io/specification#license
license: MIT

# 1-500 chars · runtime/environment requirements (omit if none)
# https://agentskills.io/specification#compatibility
# compatibility: Requires Node.js 18+. Needs network access for API calls.

# Arbitrary key-value pairs (all values coerced to strings)
# Common keys: author, version — used by Vercel, Anthropic, and community skills
# https://agentskills.io/specification#metadata
metadata:
  author: your-github-username
  version: "0.1.0"

# Space-delimited pre-approved tools (experimental — support varies by agent)
# Format: ToolName or ToolName(filter) — e.g. Bash(git:*) Bash(jq:*) Read Write
# Only include tools the skill genuinely needs; agents may ignore unknown tools
# https://agentskills.io/specification#allowed-tools
# allowed-tools: Bash(git:*) Read Write
---

<!-- BODY: loaded on activation · budget: <500 lines / ~5k tokens
     The description field controls triggering — not the body.
     The body tells the agent HOW to execute, not WHEN to activate.
     Move detailed docs to references/ and link from here.
     Write in imperative voice: "Extract the data" not "This skill extracts data"
     https://agentskills.io/specification#body -->

# my-skill-name

## Instructions

<!-- Step-by-step instructions in imperative voice.
     Be specific — agents follow instructions literally.
     Prefer structured lists and code blocks over prose. -->

## Quick Reference

<!-- Concise lookup table, decision tree, or cheat sheet.
     Agents read this after activation — make it immediately actionable.
     Example:
     | Task            | Command / Approach        |
     |-----------------|---------------------------|
     | Create widget   | Run scripts/create.sh     |
     | Update widget   | Edit assets/template.yaml | -->

## Examples

<!-- Concrete input/output examples beat explanations.
     ```
     User: "Create a dashboard widget"
     Agent: runs scripts/create-widget.sh → outputs widget.html
     ``` -->

## Constraints

<!-- Hard boundaries the agent must respect.
     - Never modify files outside the project root
     - Always validate output before returning to user -->

For extended guidelines, see [REFERENCE.md](references/REFERENCE.md).