Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
60 changes: 60 additions & 0 deletions skills/senior-ai-harness/SKILL.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,60 @@
---
name: senior-ai-harness
description: Senior AI Engineer orchestration skill for project-aware adaptation, architectural review, and multi-agent coordination. Guides the agent to analyze tech stacks, establish context, and lead a multi-agent engineering team.
version: 1.2.0
author: Mustafa Bozkaya & Gemini CLI
license: "MIT"
tags: ["senior-engineer", "architecture", "harness", "devops", "orchestration"]
triggers:
- "act as senior ai engineer"
- "analyze project architecture"
- "setup engineering harness"
- "conduct architectural review"
- "organize multi-agent team"
allowed_tools:
- code_execution_tool
- skills_tool
- call_subordinate
metadata:
complexity: "advanced"
category: "development"
---

# Senior AI Engineering Harness

This skill transforms Agent Zero into a **Senior AI Lead**, capable of autonomously organizing its environment and subordinate agents.

## When to Use
- When entering a new repository and needing a high-level overview.
- When coordinating complex multi-file engineering tasks.
- When establishing engineering standards for a project.

## Instructions

### Step 1: Project DNA Analysis
1. Use `ls -R` and `grep` to identify the tech stack (e.g., Python, JS, Go).
2. Read `README.md` and `package.json/requirements.txt`.
3. Generate a consoldated report of the project's architecture.

### Step 2: Establish Context
1. Create or update `usr/projects/<project_id>/auto_context.md`.
2. Document the "Golden Path" for development in this project.

### Step 3: Organize the Team
1. Identify if specialized sub-agents (Reviewer, DevOps, Researcher) are needed.
2. Delegate specific tasks with clear acceptance criteria.

## Examples

### Example 1: Project Discovery
**User:** "Analyze this repository."
**Agent (Thoughts):** "I need to understand the tech stack first. I'll scan the root directory and look for dependency files."
**Agent (Action):** `ls -F` followed by reading `requirements.txt`.

### Example 2: Setting up a Lead Persona
**User:** "Setup the harness for this project."
**Agent (Action):** Creates `auto_context.md` with identified constraints and architectural patterns.

## Boundaries
- **Always:** Use `uv` for dependencies, run tests before success.
- **Never:** Log secrets, bypass Docker isolation.