Learn how to expand the capabilities of CoderX and the M0X Agent with dynamic, reusable skills.
In CoderX, Skills are self-contained plugins that combine execution scripts, system prompts, API definitions, and contextual resources. By adding skills, you extend the vocabulary of tools available to the AI agent, allowing it to perform complex, domain-specific tasks (like analyzing genomics structures, fetching clinical trials, or querying custom databases).
Every skill resides in its own folder and contains at least a SKILL.md file defining its metadata and behavior:
SKILL.md (Required): The entry point containing frontmatter configuration (name, description, required packages) and detailed instructions guiding the agent on when and how to invoke the skill.scripts/: Executable Python or Shell scripts that perform actual API connections or processing tasks.resources/: static assets, reference guides, database schemas, or code templates.Instead of writing skills from scratch, you can use the Dynamic Skill Creator to package any successful terminal workflow or sub-task session into a reusable skill:
Perform your development task (e.g. configuring a complicated local database setup, running particular test scripts, or parsing logs).
Call the workflow_skill_creator tool or mention "package this workflow into a skill".
The creator analyzes your terminal command history, file diff edits, and system messages, extracting the core shell commands and instructions into a clean Python script.
A SKILL.md file is automatically generated in your local .gemini/config/plugins/ directory. You can edit the description or triggers.
SKILL.mdHere is an example structure of a custom skill manifest (SKILL.md):
---
name: db-schema-analyzer
description: Inspect local Postgres schemas and suggest indexing optimizations.
required_tools:
- run_command
- write_file
---
# Instructions for the Agent
Use this skill whenever the user asks to analyze database performance, optimize sql queries, or explain schemas.
## Step-by-Step Execution
1. Query the local Postgres schema by running the script: `python scripts/dump_schema.py`
2. Save the schema details into a temporary file.
3. Compare the indexes in the schema against common bottleneck query patterns and print recommended indexing commands in a markdown table.[!TIP] Keep your skill descriptions highly descriptive. CoderX uses the skill's description frontmatter to determine which skill is relevant to a user request during the planning phase.
C:\Users\<user>\.gemini\config\plugins\_alphafold-database).