workflow-multi-cli-plan_skill

Overview

This skill provides a unified multi-CLI collaborative planning and execution capability that routes between a full planning pipeline and a lightweight execution engine. It collects workflow preferences, enhances prompts with project context, and dispatches to the appropriate phase for plan generation or execution. The skill supports coordinated multi-agent CLI orchestration and can run in interactive or auto (no-confirm) modes.

How this skill works

The skill detects mode from the trigger (workflow:multi-cli-plan → plan, workflow:lite-execute → execute), collects user preferences, augments the prompt with available project context files, and then dispatches to the matching phase document. In plan mode it runs a multi-step planning pipeline that produces a handoff into the execution phase; in execute mode it runs a standalone execution pipeline that performs input detection, task grouping, batch execution, and review. Workflow preferences (autoYes) are passed into phases so execution can respect interactive vs automated flows.

When to use it

• You need a collaborative multi-agent CLI plan before execution across multiple CLIs or models.
• You want a lightweight execution run for a preformed task or in-memory instruction without full planning.
• You need to switch between interactive confirmations and fully automated execution.
• You want to leverage project-specific context (project-tech or guidelines) to enhance prompts.
• You need coordinated handoff from planning to execution within the same workflow.

Best practices

• Provide a clear task description or arguments when invoking the skill to improve plan quality.
• Keep .workflow/project-tech.json and .workflow/project-guidelines.json updated for richer prompt enhancement.
• Use interactive mode for initial runs to validate plan steps, then switch to Auto for repeatable runs.
• Structure tasks to allow grouping and batching so the execution pipeline can optimize runs.
• Pass exact file paths or in-memory context for execute mode to avoid ambiguous input detection.

Example use cases

• Generate a coordinated multi-CLI development plan across different language toolchains, then hand off to an execution engine.
• Run a CI-style batch execution of scripted tasks in auto mode to update dependencies and run tests.
• Perform an interactive planning session that produces step-by-step CLI commands for a complex refactor.
• Execute a one-off maintenance script provided as in-memory input without invoking the full planning pipeline.
• Switch to Lite Execute to re-run a vetted plan in fully automated mode for scheduled tasks.