home / skills / nickcrew / claude-cortex / tool-selection

tool-selection skill

safe

This skill helps you select the optimal MCP tool by evaluating task complexity and performance trade-offs with a structured decision workflow.

npx playbooks add skill nickcrew/claude-cortex --skill tool-selection

Review the files below or copy the command above to add this skill to your agents.

Files (2)

SKILL.md

997 B

---
name: tool-selection
description: Use when selecting between MCP tools based on task complexity and requirements - provides a structured selection workflow and decision rationale.
---

# Tool Selection

## Overview
Select the optimal MCP tool by evaluating task complexity, accuracy needs, and performance trade-offs.

## When to Use
- Choosing between Codanna and Morphllm
- Routing tasks based on complexity
- Explaining tool selection rationale

Avoid when:
- The tool is explicitly specified by the user

## Quick Reference

| Task | Load reference |
| --- | --- |
| Tool selection | `skills/tool-selection/references/select.md` |

## Workflow
1. Parse the operation requirements.
2. Load the tool selection reference.
3. Apply the scoring and decision matrix.
4. Report the chosen tool and rationale.

## Output
- Selected tool and confidence
- Rationale and trade-offs

## Common Mistakes
- Ignoring explicit user tool preferences
- Overweighting speed vs accuracy without justification

Overview

This skill helps pick the optimal MCP tool by evaluating task complexity, required accuracy, and performance trade-offs. It provides a structured workflow and a clear decision rationale so you can route tasks to the best tool confidently. The output includes a selected tool, confidence score, and trade-off explanation.

How this skill works

The skill parses operation requirements (latency, budget, accuracy, domain constraints) and loads a tool selection reference to establish baseline capabilities. It applies a scoring matrix that weights criteria according to priorities, compares candidate tools (for example Codanna vs Morphllm), and computes a recommendation with a confidence level. Finally, it generates a concise rationale describing chosen trade-offs and any assumptions made.

When to use it

Routing tasks between Codanna and Morphllm based on capability needs
Deciding whether to prioritize speed, cost, or accuracy for a given job
Explaining and documenting why a specific tool was selected
Automating pipeline decisions in agent orchestration or CLI workflows
Evaluating tool suitability for enterprise constraints or domain-specific tasks

Best practices

Always respect explicit user tool preferences and override automatic selection if specified
Define and weight selection criteria (latency, accuracy, cost, robustness) before scoring
Log the decision matrix and assumptions to enable auditability and future tuning
Re-evaluate weights periodically as tool performance or costs change
Avoid biasing scores toward speed or cost without documenting the trade-off against accuracy

Example use cases

Selecting Morphllm when high accuracy on domain-specific data outweighs latency concerns
Choosing Codanna for low-latency, high-throughput tasks where approximate results are acceptable
Routing customer support classification to the tool with the best proven accuracy for that language
Automating task assignment in a TUI or CLI that manages mixed workloads
Producing a human-readable recommendation report for engineering review or compliance

FAQ

What if a user explicitly requests a specific tool?

Honor the explicit request and skip automatic selection; include the user preference in logs and rationale.

How is confidence calculated?

Confidence is derived from the normalized score difference between top candidates and the completeness of required capability matches.