# Metacognitive Plan

> Orchestrates the LSTR 6-step metacognitive reasoning chain and produces a decision-complete plan against verifiable constraints. Invoke for architecture decisions, multi-constraint trade-offs, incident triage, or any task requiring auditable reasoning. Trigger phrases "plan this", "/plan" (delegating to the reasoning chain), "apply the LSTR framework", "reason through this", or any query where constraints compete and a plan is needed.

- Skill: `lstr-1/metacognitive-plan` (Agent Skill)
- Install (CLI): `npx skillmds@latest add lstr-1/metacognitive-plan`
- Raw SKILL.md: https://api.skillmd.com/api/skills/lstr-1/metacognitive-plan/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: LSTR-1 (https://skillmd.com/u/lstr-1)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/lstr-1/metacognitive-plan

---


# metacognitive-plan

Runs the LSTR 6-step chain (delegating the exact tool-argument contract to the global `lstr-reasoning-framework` skill), then adds constraint extraction and mental simulation before committing to a plan.

## When to Use

- Multi-step inference, competing constraints, non-trivial uncertainty, or material consequence (≥2 complexity-gate criteria).
- Any query where an auditable reasoning trail is required.

## Instructions

1. **Complexity gate** — evaluate against `10-reasoning-framework.md`. If <2 criteria, skip the chain and state "framework not warranted".
2. **Run the 6-step chain** via the lstr-reasoning MCP family, per `lstr-reasoning-framework`:
   - `metacognitiveMonitoring` (knowledge-assessment) → `sequentialthinking` (≥3 thoughts) → `collaborativeReasoning` (≥2 divergent personas) → `scientificMethod` + `structuredArgumentation` (thesis + objection + synthesis) → `constraintSolver` (numeric only).
   - EXCLUDE `narrative-planner`.
3. **Constraint extraction (DeepPlanning)** — classify every constraint as `local` (step-scoped) or `global` (cross-subtask budget / dependency). Record each constraint's verifiability: deterministic, rule-checkable, or qualitative.
4. **Mental simulation (Qwen-AgentWorld)** — before committing, rehearse the top candidate plan against the environment model; predict where it fails under resource limits or silent degradation; backtrack if a global constraint is violated.
5. **Emit a decision-complete plan** — ordered, dependency-aware steps, each with an exit condition. No calendar dates; use predecessor relationships.

## Output Contract

- Decision-complete plan (executor makes no additional choices).
- Local vs global constraint list, each with verifiability label.
- Reasoning record persisted per `lstr-reasoning-framework` contract.
- Confidence from `metacognitiveMonitoring.overallConfidence`.

## Constraints

- Do NOT fabricate a reasoning trace. A failed step is logged `failed`.
- Do NOT encode qualitative constraints as `constraintSolver` booleans.