← Về thư mục
📄 / / root / GitHub / .agents / workflows / atomic-complexity-wf.md

description: "Workflow for automatically breaking complex rules/skills into granular, task-specific modules."

// turbo-all

Atomic Complexity Workflow

This workflow is triggered whenever a task involves multiple technical layers (API, UI, Logic, DB) OR when a Rule (.mdc), Skill (SKILL.md), or Workflow (.md) file exceeds the Cognitive Threshold (approx. 100 lines or 15 distinct rules).

1. Modular Decomposition

Identify the "Instruction Contexts" that can be decoupled: - Technical/Syntax: Regex patterns, ASCII formats, Specific API key formats. - Domain Logic: Astrological calculation rules, Dasha sequences. - Interactions: UI component behavior, Event hooks. - External: API integration protocols.

2. The Granular Split

If a file or task contains more than TWO of the above contexts, it MUST be split into smaller, task-specific modules: - Move "Technical/Syntax" into a dedicated Technical Skill (e.g., vedic_parser). - Keep "Core Logic" in a Logic Skill (e.g., vedic_logic). - Store UI/Interactions in a UI Rule or Component Skill.

3. Reference & Hierarchy

Maintain a "Main Skill" or "Master Rule" that links to these sub-modules at the end. - Use file-protocol links: ### [Parsing Rules](file:///Users/keira/Documents/GitHub/Vedic-astro-mar/.agents/skills/vedic_parser/SKILL.md) - This prevents AI context pollution by allowing the agent to "drill down" only when necessary.

4. Mandatory Approval Step

CRITICAL: Every "Auto-Breaking" operation MUST generate a structured Implementation Plan for user approval before moving any files or logic.

The plan must follow the format defined in .cursor/rules/implementation-plans.mdc.