mcpField
G
goal-plan

goal-plan

github.com/ruvnetAgent workflows80.0k770 installsUpdated today
Repository

About goal-plan

A Ruflo skill for planning complex objectives with Goal-Oriented Action Planning. It models action prerequisites, outcomes, and costs, then tracks execution and replans when conditions change.

Key features

  1. 01Model goals and current state
  2. 02Define action prerequisites and effects
  3. 03Generate action sequences with A*
  4. 04Track task execution trajectories
  5. 05Replan after failures or changes
  6. 06Store successful plans in memory

Use cases

  1. 01Plan tasks with interdependent steps
  2. 02Adapt plans after unexpected outcomes

MCP support

Uses Ruflo MCP tools to track trajectories, create tasks, and store successful plans.

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/ruvnet/ruflo --skill goal-plan

After installing, type/goal-planin Claude Code to confirm it loaded.

Details

From the project README

Goal Plan

Create and execute intelligent plans using Goal-Oriented Action Planning (GOAP).

When to use

When you have a complex objective that requires multiple steps, has dependencies between steps, and may need adaptive replanning as conditions change.

Steps

  1. Define goal state — what does "done" look like? List concrete success criteria
  2. Assess current state — what's true now? What assets, code, infrastructure exist?
  3. Identify gap — what must change between current and goal state?
  4. Inventory actions — list available actions with:
    • Preconditions (what must be true before this action)
    • Effects (what becomes true after this action)
    • Cost estimate (time, complexity, risk)
  5. Generate plan — find the optimal action sequence using A* through the state space
  6. Record trajectory — call mcp__plugin_ruflo-core_ruflo__hooks_intelligence_trajectory-start to begin tracking
  7. Create tasks — call mcp__plugin_ruflo-core_ruflo__task_create for each action in the plan
  8. Execute — work through tasks in dependency order:
    • Before each action: verify preconditions still hold
    • After each action: verify effects achieved
    • Record each step via mcp__plugin_ruflo-core_ruflo__hooks_intelligence_trajectory-step
  9. Monitor & replan — if an action fails or produces unexpected results:
    • Reassess current state
    • Recalculate optimal path from new state
    • Update remaining tasks
  10. Complete trajectory — call mcp__plugin_ruflo-core_ruflo__hooks_intelligence_trajectory-end
  11. Store successful plan — call mcp__plugin_ruflo-core_ruflo__memory_store with namespace goap-plans

Plan output format

Goal: [concrete objective]
Current State: [key facts]
Plan Cost: [estimated effort]
Steps:
  1. [action] — precondition: [X], effect: [Y], cost: [Z]
  2. [action] — precondition: [Y], effect: [W], cost: [Z]
  ...
Risk Factors: [what could force a replan]
Fallback: [alternative approach if primary path fails]

Replanning triggers

  • Action fails (precondition no longer met)
  • Unexpected side effects detected
  • New information changes goal definition
  • Cost exceeds threshold
  • External dependency becomes unavailable

Comments

Markdown supported