--- name: rules-workflows description: Use to standardize project context (Rules) or automate complex multi-step tasks (Workflows). license: MIT compatibility: Requires Antigravity CLI allowed-tools: read_file write_to_file run_command --- # Rules & Workflows Orchestrator This skill outlines the standards for acting as an **Autonomous Process Orchestrator**, managing behavior and workflows to align with Agile best practices and project constraints. ## 🧠 Core Philosophy: Autonomy & Reasoning **"Think before you Act"** Before creating any rule or workflow, the Agent MUST: 1. **Analyze the Goal**: What is the user trying to achieve? 2. **Orchestrate Capabilities**: Identify which _other_ skills (Frontend, Backend, QA) are needed. 3. **Separate Concerns**: - **Workflows** = Process (Steps, Sequences) - **Skills** = Knowledge (Standards, Implementation) ## 📚 Reference Library This skill relies on specialized guides located in the `references/` directory. **You MUST consult these files for specific tasks.** | Task | Reference File | Purpose | | :---------------------------- | :------------------------------------- | :------------------------------------------------------------------------------ | | **Creating & Managing Rules** | `references/rules-guide.md` | Standard Operating Procedures (SOPs), Context Injection, Rule Activation Types. | | **Building Workflows** | `references/workflows-guide.md` | High-Performance Workflow Design, "Turbo" Execution, Checkpoints. | | **Orchestrating Skills** | `references/orchestration-patterns.md` | Skill Chaining Patterns (e.g., Spec-First, TDD), Conflict Resolution. | --- ## 1. Orchestration Strategy **👉 For best practices on sequencing skills, refer to `references/orchestration-patterns.md`**. Key principles: 1. **Never Silo Skills**: Code requires Architecture; Design requires Requirements. 2. **Sequential Thinking**: Use `sequential-thinking` for complex problems before acting. 3. **Process vs. Knowledge**: Ensure workflows delegate "how-to" knowledge to the appropriate Skill. --- ## 2. Self-Correction & Learning The Agent can modify its own rules (Meta-Programming). - **User Correction**: "Don't do X anymore" -> Trigger `workflow-rule-from-feedback.md`. - **New Project**: "Read this codebase" -> Trigger `workflow-rule-from-codebase.md`.