stegoro — Auto-generate steering policies for AI agents

Why

Building a purpose-built AI agent is hard. Without a structured policy, agents produce inconsistent outputs, skip validation, and lack quality assurance. Writing these policies by hand is time-consuming and error-prone.

stegoro automates the entire policy creation process — from understanding the agent's purpose to generating a complete, quality-calibrated policy set.

What

4-Phase Workflow — Discovery, Design, Generation, and Refinement
Adaptive Depth — Adjusts research depth based on domain complexity
4 Agent Types — Process, Task, Analytical, and Hybrid patterns
AI-DLC Quality — Validates against 11 quality dimensions
Interactive Dialogue — Approval gates at every critical decision
Session Continuity — Pause and resume across sessions
Complete Audit Trail — ISO 8601 timestamped decision logging

How It Works

Phase	Purpose	Key Activities
Discovery	Understand what & why	Purpose analysis, domain research, stakeholder identification, scope definition
Design	Architect the policy	Workflow architecture, common rules, phase rules, quality mechanisms
Generation	Create policy files	Core workflow, common rules, phase rules, integration validation
Refinement	Ensure quality	Completeness review, consistency review, quality calibration

Getting Started

Prerequisites: Claude Code installed and configured.

1. Add the marketplace (one-time):

/plugin marketplace add yoshidashingo/steering-policy-generator

2. Install the plugin:

/plugin install stegoro@steering-policy-generator

3. Start generating:

/stegoro:generate a code review agent

4. Follow the interactive 4-phase workflow. The generator never advances without your explicit approval.

FAQ

What types of agents can it generate policies for?: Any purpose-built AI agent. It supports Process agents (multi-phase workflows), Task agents (focused operations), Analytical agents (data-driven), and Hybrid agents (mixed patterns).
How long does it take to generate a policy?: It depends on domain complexity. A simple task agent may take one session; a complex process agent with multiple stakeholders may span several sessions.
Can I customize the generated policy?: Yes. The generator produces editable Markdown files. You can modify any file after generation, and the Refinement phase helps ensure quality of modifications.
What is AI-DLC?: AI-Driven Development Life Cycle. It defines 11 quality dimensions used as the benchmark for generated policies, including adaptive workflow, mandatory checkpoints, audit trail, and overconfidence prevention.