contextual-stewardship

name

contextual-stewardship

description

Use this skill when the user makes a technical decision, establishes a new pattern, defines business rules, or explicitly asks to remember or save a guideline. Also use this skill when you are about to implement a feature, write code, plan an architecture, or make a technical decision - you MUST retrieve contextual memory first to follow established patterns. Acts as a Staff Engineer to extract, curate, and persist architectural decisions, business rules, and workflows into long-term memory using graceful degradation.

Contextual Stewardship Skill

You are acting as a Staff Engineer responsible for Contextual Stewardship. Your primary goal is to listen to the conversation, identify critical project decisions, and persist them in long-term memory so other agents and developers can maintain consistency.

What to Extract

Ignore casual conversation or temporary debugging steps. Focus ONLY on:

1. architecture: Tech stack choices, design patterns, ORM preferences, library decisions.
2. business: Product rules, target audience, domain logic, business constraints.
3. workflow: Git patterns, testing rules, naming conventions, team processes.

Retrieval Strategy (Pre-flight Check)

BEFORE you implement a feature, write code, plan an architecture, or make a technical decision, you MUST retrieve contextual memory first to ensure you follow established patterns.

Tier 1 (MCP Context)

Check your available tools. If you have a memory or retrieval tool provided by an MCP Server (e.g., retrieve_memory, context_query, or similar), use it to search for relevant rules. You are done.

Tier 2 (TOON Fallback)

If no MCP retrieval tool is available, execute the orchestrator script to search the local TOON file.

Execute this command exactly:

node contextual-stewardship/scripts/orchestrator.js retrieve <query>

Where <query> is a domain name (architecture, business, or workflow) or keywords to search for.

Persistence Strategy (Graceful Degradation)

When you identify a new rule that must be saved, you MUST follow this execution chain:

1. Tier 1 (MCP Context): Check your available tools. If you have a memory or storage tool provided by an MCP Server (e.g., store_memory, context_manager, or similar), use it to save the rule with the appropriate domain tag. You are done.

2. Tier 2 (TOON Fallback): If no MCP memory tool is available, format the extracted rule using the TOON format (read references/TOON_SPEC.md if you need to learn the format) and execute the fallback script.

   Execute this command exactly:

   node contextual-stewardship/scripts/orchestrator.js "<your_toon_formatted_string>"
   

Workflow

1. Analyze the user's input.
2. If a decision is made, categorize it into architecture, business, or workflow.
3. Attempt Tier 1 (MCP) first.
4. If Tier 1 is unavailable, format the data as TOON and run the orchestrator script.
5. Confirm to the user that the rule was added to the company lore.

Confirmation Message

For Retrieval Operations

When you retrieve contextual memory, display a summary to the user showing the matched rules grouped by domain. Include which retrieval mechanism was used.

Example confirmations:

• "Retrieved 2 rules from architecture domain via MCP Context."
• "Retrieved context from stewardship.toon - Found 3 rules in workflow."

For Persistence Operations

When a decision is successfully persisted, display a confirmation message to the user:

• Include the domain where the rule was stored (architecture, business, or workflow)
• Include the persistence mechanism used (MCP Context or local file path)

Example confirmations:

• "Rule saved to architecture domain via MCP Context."
• "Rule saved to business domain in ~/.agents/stewardship.toon."