NavanithanS

ask-skill-creator

Use this skill when the user asks to create a new skill or add a new capability to the Agent Skill Kit. Triggers: "create a skill", "new skill", "add skill for", "teach the agent". Do NOT use this skill for: - Creating normal software features (use a coding architect). - Modifying the core ASK runtime. Capabilities: - Scaffolds the "Gold Standard" directory structure (SKILL.md, scripts/, assets/, tests/). - Generates optimized Frontmatter and Algorithmic Instructions. - Updates the root README.md.

NavanithanS 1 1 Updated 3mo ago

Resources

4
GitHub

Install

npx skillscat add navanithans/agent-skill-kit/ask-skill-creator

Install via the SkillsCat registry.

SKILL.md

Skill Creation Protocol

  1. NO generic descriptions. Be highly specific and semantic.
  2. NO missing directories. MUST create scripts/, assets/, tests/.
  3. MUST prefix skill name with ask- and use kebab-case.
  4. MUST include <critical_constraints>, <process>, and <thinking> sections in the new SKILL.md.</critical_constraints>

skills///
├── SKILL.md # 🧠 The Brain (Frontmatter + Protocol)
├── scripts/ # 🛡️ Guardrails (Empty .py/.js placeholders)
├── assets/ # 📚 Knowledge (examples.md)
└── tests/ # 🧪 Verification (case1.md)
</file_structure>

  1. Analyze Request:

    • Determine category (coding, planning, tooling).
    • Determine skill-name (must start with ask-).
  2. Template Design:

    • Draft the SKILL.md frontmatter with clear Triggers and Negative Constraints.
    • Design 3-5 algorithmic steps for the <process> section.
    • Plan at least one "Critical Constraint" specific to this skill.
  3. Scaffold Directory:

    • Create skills/<category>/<skill-name>/.
    • Create subfolders: scripts/, assets/, tests/.
  4. Generate Content:

    • SKILL.md: Write the full protocol.
    • assets/examples.md: Create a placeholder or basic example.
    • tests/case1.md: Create a sample input/output test case.
  5. Register:

    • Update the root README.md skill table with the new skill.

SKILL.md Frontmatter

---
name: ask-example-skill
description: >
  Use when [user intent].
  Triggers: "phrase 1", "phrase 2".
  
  Do NOT use for:
  - [Constraint 1]
---

SKILL.md Body

# [Skill Name] Protocol

## <critical_constraints>
1. ...
</critical_constraints>

## <process>
1. **Step 1**: ...
2. **<thinking> Plan**: ...
   </thinking>
3. **Step 3**: ...
</process>