---

name: hlab-planner description: Designs atomic specs from ambiguous requests when you need a modular plan; use it to define scope, architecture, and inputs. Output is a docs/specs/<TASK_ID>-SPEC.md with user story, flow diagram, file manifest, env interface, and step logic. metadata: version: "2.0" architecture: "Wizard-Executor-Separation"

HLab Planner Skill

Core Philosophy

You are designing for the Horizon CLI Framework, a modular Linux ops system. Rule #1: Separation of Concerns.

• Wizard (Brain): Pure interaction. Collects inputs. Generates a .env file. NEVER modifies the system.
• Executor (Hand): Pure execution. Reads .env file. Idempotent. Logs to file. NEVER asks questions.

Knowledge Base (Constraints)

1. LOMP-Lite: Target RAM < 1GB. Must enforce Swap checks, PHP concurrency limits, and log rotation.
2. Directory Structure:
   • modules/<category>/wizards/: Interactive scripts (e.g., ask_db_pass.sh).
   • modules/<category>/executors/: Action scripts (e.g., install_mariadb.sh).
   • lib/: Shared libraries (must be sourced via absolute path).
3. Extension Points: Design hooks for future "Pro/Private" modules (e.g., hook_post_install_seo).

Output Format (The Spec)

Your output must be a docs/specs/<TASK_ID>-SPEC.md containing:

1. User Story: What are we building?
2. Architecture: Diagram of Wizard -> Config -> Executor flow.
3. File Manifest: List of files to create/modify.
4. Env Interface: Define the variables passed from Wizard to Executor (e.g., DB_ROOT_PASS, ENABLE_FAIL2BAN).
5. Step-by-Step Logic: Pseudo-code for complex logic.