Skill Invocation Trace 🔍

A specialized skill for Gemini CLI, Claude Code, and other agents to provide transparent, observable, and auditable skill invocation traces.

🌟 Why use this?

Have you ever wondered why your AI agent chose a specific skill, or what files it actually read to perform a task?

skill-invocation-trace makes the "magic" observable by generating a structured report of:

• Trigger Evidence: Why the skill was activated.
• Resource Inventory: Every SKILL.md, reference file, and script involved.
• Execution Plan: The step-by-step logic the agent followed.

🚀 Installation

You can install this skill globally using the standard npx skills CLI:

npx skills add kkxjerry/skill-invocation-trace -g

🛠 Usage

Once installed, you can ask the agent to explain its behavior using the following prompt:

"Use $skill-invocation-trace to explain why this skill triggered and how it will be used for the current task."

Example Output

Skill invocation trace

Target skill
- meeting-minutes

Why it triggered
- Observed: the user provided a meeting transcript and asked for a summary.
- Inferred: matches the core intent of the meeting-minutes skill.

Files read
- ~/.agents/skills/meeting-minutes/SKILL.md

Execution plan
1. Read the provided transcript.
2. Extract key participants and action items.
3. Format output as a Markdown summary.

📂 Structure

• SKILL.md: Core logic and system instructions.
• agents/: Agent-specific configuration (OpenAI/Gemini/Claude).
• references/: Detailed guides on observability limits.
• scripts/: Helper scripts for automated trace rendering.

🤝 Contributing

Feel free to open issues or submit pull requests if you have ideas to make agent behavior even more transparent!

---

Created with ❤️ for the Agent Skills ecosystem.