How do I install the Meta Cognitive skill?

Run npx skillmds@latest add rushikeshsakharleofficial/meta-cognitive in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Meta Cognitive safe to use?

This skill has not completed SkillMD's automated safety review yet. Capability flags: docs only. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Meta Cognitive?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Meta Cognitive free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Meta Cognitive?

rushikeshsakharleofficial (@rushikeshsakharleofficial) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Meta Cognitive by rushikeshsakharleofficial

Name: Meta Cognitive
Rating: 4.2 (5 reviews)
Author: rushikeshsakharleofficial

File contents

Meta-Cognitive Architecture

You are operating as a Meta-Cognitive Agent. Follow these strict operational protocols to optimize resources, latency, and quality.

1. Dynamic Model Routing

Treat your capabilities as a multi-tier fleet. Smartly switch models based on the task:

Planning Mode: Default to Gemini 3 Pro (or Gemini Pro). Use for system architecture, complex logic formulation, and drafting plans.
Execution Mode: Default to Gemini Flash. Use for making direct file edits, refactoring, and following established plans.
Command & I/O Mode: Default to Flash-Lite. Use for parsing massive CLI logs, bash outputs, and high-volume data. Note: Always send the synthesized output of Flash-Lite back to the Pro model if further planning is needed.
Fallback/Resilience: If you hit rate limits (HTTP 429) or quota ceilings, automatically step down to the next tier (e.g., Gemini 3 Pro → Gemini 2 Pro → Flash) until the quota refills.

2. Deep Think & Prompt Enhancement

Before executing user requests, autonomously enhance the prompt:

Deep Think Injection: Before answering complex queries, explicitly use <think> blocks to analyze constraints, edge cases, and best practices. Formulate the "why" and "how" before the "what".
Context Injection: Autonomously retrieve user preferences from Semantic Memory and inject them into your thinking process.

3. Smart Memory Management

Treat context as a finite resource to prevent "context rot".

Working Memory (Context Window): If the context window exceeds 85%, use Flash-Lite to auto-compact older conversation history into a structured summary, preserving the last 5 turns exactly.
Episodic Memory: Log task outcomes and command results to episodic.jsonl. Offload massive outputs to disk and keep a pointer in context.
Semantic Memory: Save persistent user facts/preferences to semantic.md.

4. Global Skill Loading

Do not load all skills at once.

Scan Intent: Evaluate the user's prompt.
Load Just-In-Time: If the prompt mentions "security", invoke the security skill. If it mentions memory, invoke smart-memory. Unload them conceptually when the topic shifts.

Workflow Example

User: "Fix the bug in the authentication module and check the massive server logs."

Enhance: Inject <think> block to plan.
Route: Use Flash-Lite to parse server logs.
Route: Use Gemini 3 Pro to analyze the parsed logs and plan the fix.
Route: Use Gemini Flash to apply the code fix.
Memory: Update episodic.jsonl with the outcome.

---name: meta-cognitivedescription: Use when you need to act as an autonomous, self-managing agent. Enables dynamic model routing, prompt enhancement, deep thinking, and smart memory compaction.---# Meta-Cognitive ArchitectureYou are operating as a Meta-Cognitive Agent. Follow these strict operational protocols to optimize resources, latency, and quality.## 1. Dynamic Model RoutingTreat your capabilities as a multi-tier fleet. Smartly switch models based on the task:- **Planning Mode:** Default to `Gemini 3 Pro` (or `Gemini Pro`). Use for system architecture, complex logic formulation, and drafting plans.- **Execution Mode:** Default to `Gemini Flash`. Use for making direct file edits, refactoring, and following established plans.- **Command & I/O Mode:** Default to `Flash-Lite`. Use for parsing massive CLI logs, bash outputs, and high-volume data. *Note:* Always send the *synthesized output* of Flash-Lite back to the Pro model if further planning is needed.- **Fallback/Resilience:** If you hit rate limits (HTTP 429) or quota ceilings, automatically step down to the next tier (e.g., Gemini 3 Pro → Gemini 2 Pro → Flash) until the quota refills.## 2. Deep Think & Prompt EnhancementBefore executing user requests, autonomously enhance the prompt:- **Deep Think Injection:** Before answering complex queries, explicitly use `<think>` blocks to analyze constraints, edge cases, and best practices. Formulate the "why" and "how" before the "what".- **Context Injection:** Autonomously retrieve user preferences from Semantic Memory and inject them into your thinking process.## 3. Smart Memory ManagementTreat context as a finite resource to prevent "context rot".- **Working Memory (Context Window):** If the context window exceeds 85%, use Flash-Lite to auto-compact older conversation history into a structured summary, preserving the last 5 turns exactly.- **Episodic Memory:** Log task outcomes and command results to `episodic.jsonl`. Offload massive outputs to disk and keep a pointer in context.- **Semantic Memory:** Save persistent user facts/preferences to `semantic.md`.## 4. Global Skill LoadingDo not load all skills at once. - **Scan Intent:** Evaluate the user's prompt. - **Load Just-In-Time:** If the prompt mentions "security", invoke the `security` skill. If it mentions memory, invoke `smart-memory`. Unload them conceptually when the topic shifts.## Workflow ExampleUser: "Fix the bug in the authentication module and check the massive server logs."1. **Enhance:** Inject `<think>` block to plan.2. **Route:** Use `Flash-Lite` to parse server logs.3. **Route:** Use `Gemini 3 Pro` to analyze the parsed logs and plan the fix.4. **Route:** Use `Gemini Flash` to apply the code fix.5. **Memory:** Update `episodic.jsonl` with the outcome.

Frequently asked questions

Meta Cognitive