AI Agent Skill

The skills/cloud-fs directory contains an installable AI agent skill that teaches coding assistants (Claude Code, Gemini CLI, etc.) how to use cloud-fs as a POSIX-like virtual filesystem.

Why use a Skill?

While the MCP server provides the tools, the Skill provides the intuition. It teaches the LLM:

• How to map high-level user intents ("show me my bucket") to specific tool calls (list_directory).
• How to handle first-time setup (bootstrap flow) if the MCP server isn't configured yet.
• How to use the server's unique features like line-range reads and regex-based search.

Installation

Claude Code

claude mcp add-skill nogoo9/mcp-server-cloud-fs

Gemini CLI (npx)

npx skills add nogoo9/mcp-server-cloud-fs

Gemini CLI (local)

If you have the repository cloned locally:

bun x skills add ./skills/cloud-fs

Features

MCP Mode Auto-detection

The skill automatically recognizes when cloud-fs is configured in the environment (e.g., via .mcp.json) and tells the assistant how to use the mcp__cloud-fs__* tools.

Bootstrap Flow

If the assistant detects that cloud-fs is needed but not yet configured, the skill provides a guided setup flow. It will ask the user for:

1. The cloud provider (S3, Azure, GCS, etc.)
2. The root URI (e.g., s3://my-bucket)
3. Credentials (as environment variables)

It then offers to save this configuration to a local .mcp.json file for future sessions.

POSIX-to-MCP Mapping

The skill includes a comprehensive mapping that translates standard shell commands into the most efficient MCP tool calls:

Intent / Command	MCP Tool Call
ls, dir	list_directory
cat, less	read_file or read_file_range
grep	grep_files
find	search_files
cp, mv, rm	move_file, delete_file
mkdir, touch	write_file

Provider Credential Guidance

The skill acts as a living reference for how to configure credentials for various providers (AWS IAM, Azure Managed Identity, GCS Service Accounts), reducing the need for the user to consult external documentation.