MCP SSH Manager

This skill should be used when the user asks to "run ssh command", "execute on server", "ssh session", "upload file", "download file", "ssh tunnel", "check server status", "monitor server", "deploy files", "backup server", or needs remote server management. This skill emphasizes session reuse, workdir organization, and content persistence for sustainable operations.

Security Scan

The skill claims to manage SSH sessions and files but omits how the required tooling/credentials are provided and instructs the agent to persist potentially sensitive command outputs to disk, creating several coherence and data‑exposure risks you should review before installing.

Purpose & Capability

The skill's name and description match SSH management. However, the runtime instructions reference custom CLI tools (ssh_execute, ssh_session_start, ssh_upload, etc.) and standard operations (ssh, scp, tar, git, pm2) without declaring any required binaries or credentials. The package contains helper scripts for creating a workdir and logging, but does not provide or install the external 'mcp ssh-manager' server or explain how the ssh_* tools are made available. That mismatch (claims to provide/manage servers but no install, no required binaries, no primary credential) is incoherent.

Instruction Scope

SKILL.md instructs the agent to run arbitrary remote commands, create persistent sessions, transfer files, and save command outputs and host snapshots under ~/.ssh-workdir. Those behaviors are expected for an SSH manager, but they also permit capturing and storing arbitrary remote output (which may include secrets), and running any command on remote hosts. The instructions are broad and rely on external tooling and existing SSH auth; they also implicitly assume access to the user's SSH agent/keys.

Install Mechanism

There is no install spec (instruction-only), which minimizes arbitrary remote downloads. The repo includes three small scripts (create-workdir.sh, log-command.sh, save-status.sh) that manage local logs/workdirs. Not having an install step reduces installer risk, but it increases ambiguity about where the referenced ssh_* tools come from.

Credentials

The skill declares no required environment variables or credentials, yet its functionality implicitly requires SSH authentication (private keys, agent, or credentials) and uses standard system tools. The lack of explicit credential requirements is surprising: the skill will depend on the user's existing SSH keys/config and will store outputs locally. This implicit access should be made explicit so users know what will be used or exposed.

Persistence & Privilege

The skill persists command logs, outputs, and status snapshots under ~/.ssh-workdir. Persisting remote outputs locally is reasonable for auditing, but it increases the risk of storing sensitive data (passwords, tokens, config) without explicit safeguards. The skill does not set always:true, but also does not disable model invocation — meaning the agent could invoke these behaviors when the skill is eligible. Combined with the ability to run arbitrary remote commands, that is a noteworthy privilege.

What to consider before installing

Before installing, verify the skill's source and how the ssh_* tools are provided (is there an external 'mcp-ssh-manager' service you must run?). Review the three included scripts to ensure they don't exfiltrate data or run unexpected network calls. Be aware the skill will write command outputs and host snapshots to ~/.ssh-workdir — inspect those outputs regularly and avoid storing sensitive data there. Confirm how SSH authentication will occur (SSH keys/agent), and consider restricting model-triggered use (disableModelInvocation) or running the skill in a sandbox until you trust it. If you can't verify the upstream project/homepage or the origin of the ssh_* helpers, treat this package as untrusted.

MCP SSH Manager Skill

Original MCP Server: mcp-ssh-manager by @bvisible

This skill provides documentation, workflows, and best practices for using the MCP ssh-manager server.

Manage remote SSH servers using MCP ssh-manager tools. Emphasizes session reuse, workdir organization, and content persistence for sustainable operations.

Quick Reference

Connection Management

Task	Tool	Example
List servers	ssh_list_servers	ssh_list_servers
Execute command	ssh_execute	ssh_execute server="rock5t" command="df -h"
Execute with sudo	ssh_execute_sudo	ssh_execute_sudo server="rock5t" command="apt update"
Check status	ssh_connection_status	ssh_connection_status action="status"

Session Management

Task	Tool	Example
Start session	ssh_session_start	ssh_session_start server="rock5t" name="deploy"
Send command	ssh_session_send	ssh_session_send session="xxx" command="cd /var"
List sessions	ssh_session_list	ssh_session_list
Close session	ssh_session_close	ssh_session_close session="xxx"

File Operations

Task	Tool	Example
Upload file	ssh_upload	ssh_upload server="rock5t" localPath="." remotePath="/tmp"
Download file	ssh_download	ssh_download server="rock5t" remotePath="/var/log/syslog" localPath="."
Sync files	ssh_sync	ssh_sync server="rock5t" source="local:./dist" destination="remote:/var/www"

Monitoring

Task	Tool	Example
Tail log	ssh_tail	ssh_tail server="rock5t" file="/var/log/syslog" lines=20
Health check	ssh_health_check	ssh_health_check server="rock5t"
Monitor resources	ssh_monitor	ssh_monitor server="rock5t" type="overview"
Service status	ssh_service_status	ssh_service_status server="rock5t" services="nginx,docker"

Tunneling

Task	Tool	Example
Create tunnel	ssh_tunnel_create	ssh_tunnel_create server="rock5t" type="local" localPort=8080 remoteHost="localhost" remotePort=80
List tunnels	ssh_tunnel_list	ssh_tunnel_list
Close tunnel	ssh_tunnel_close	ssh_tunnel_close tunnelId="xxx"

Backup

Task	Tool	Example
Create backup	ssh_backup_create	ssh_backup_create server="rock5t" type="files" name="data"
List backups	ssh_backup_list	ssh_backup_list server="rock5t"
Restore backup	ssh_backup_restore	ssh_backup_restore server="rock5t" backupId="xxx"
Schedule backup	ssh_backup_schedule	ssh_backup_schedule server="rock5t" schedule="0 2 * * *" type="files" name="daily"

Usage Examples

Example 1: Single Command

# Simple command - no session needed
ssh_execute server="rock5t" command="df -h"

Example 2: Multi-step Deployment with Session

# Check existing sessions first
ssh_session_list

# Start a persistent session
ssh_session_start server="rock5t" name="deploy"

# Get session ID from previous response
ssh_session_send session="xxx" command="cd /home/imax/project"
ssh_session_send session="xxx" command="git pull origin main"
ssh_session_send session="xxx" command="npm install"
ssh_session_send session="xxx" command="npm run build"
ssh_session_send session="xxx" command="pm2 restart all"

# Close when done
ssh_session_close session="xxx"

Example 3: System Health Check

# Check overall health
ssh_health_check server="rock5t"

# Monitor specific resources
ssh_monitor server="rock5t" type="cpu" interval=5 duration=30

# Check specific services
ssh_service_status server="rock5t" services="nginx,docker,postgres"

Example 4: File Deployment

# Upload deployment package
ssh_upload server="rock5t" localPath="./dist/app.tar.gz" remotePath="/tmp/app.tar.gz"

# Extract and restart
ssh_execute server="rock5t" command="cd /tmp && tar -xzf app.tar.gz && cp -r app/* /var/www/ && pm2 restart app"

Example 5: Log Monitoring

# Tail real-time logs
ssh_tail server="rock5t" file="/var/log/nginx/access.log" lines=50 follow=true

# Filter with grep
ssh_tail server="rock5t" file="/var/log/syslog" grep="error" lines=100

Example 6: Create SSH Tunnel

# Local port forward (access remote service locally)
ssh_tunnel_create server="rock5t" type="local" localPort=5432 remoteHost="localhost" remotePort=5432

# Now connect to local:5432 to access remote database

Workdir Management

Store SSH operation results in ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/ for reuse and comparison.

Structure

~/.ssh-workdir/
└── {hostname}/
    └── {YYYY-MM-DD}-{topic}/
        ├── commands.md    # All executed commands
        ├── output/        # Command outputs
        │   ├── df-h.txt
        │   ├── cpu.txt
        │   └── memory.txt
        ├── status.json    # Host status snapshot
        └── summary.md     # Findings and notes

Create Workdir

# Create new workdir
mkdir -p ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/output

# Create commands log
touch ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/commands.md

Log Commands

# Add command to log
echo "## $(date)" >> ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/commands.md
echo 'df -h' >> ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/commands.md

Save Output

# Execute and save
ssh_execute server="{hostname}" command="df -h" > ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/output/df-h.txt

Write Summary

# Write findings
echo '## System Check Findings' >> ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/summary.md
echo '- Disk usage: 75% on /dev/sda1' >> ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/summary.md
echo '- Memory: 4GB/16GB used' >> ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/summary.md

Reuse Previous Context

# Check if recent work exists
ls ~/.ssh-workdir/{hostname}/

# Read previous summary
cat ~/.ssh-workdir/{hostname}/{previous-date}-{topic}/summary.md

# Compare outputs
diff ~/.ssh-workdir/{hostname}/{yesterday}-{topic}/output/df-h.txt \
     ~/.ssh-workdir/{hostname}/{today}-{topic}/output/df-h.txt

Session Management Guidelines

When to Use Sessions

Use session for:

• Multi-step deployments
• Tasks requiring state (cd, environment)
• Long-running workflows (more than 3 commands)
• Tasks where command order matters

Don't use session for:

• Single quick commands (df -h, pwd)
• Unrelated commands that don't need state
• Read-only monitoring tasks

Session Lifecycle

# 1. Check existing sessions first
ssh_session_list

# 2. Reuse existing session if available and still active
ssh_session_send session="existing-id" command="..."

# 3. Start new session only if necessary
ssh_session_start server="{hostname}" name="{task-name}"

# 4. ALWAYS close when done
ssh_session_close session="{session-id}"

Timeout Considerations

• SSH server may close idle sessions (typically 3-5 minutes by default)
• Configure ClientAliveInterval on server for longer keepalive
• For long-running tasks, consider periodic lightweight commands to keepalive
• If session becomes unresponsive, create a new one

Best Practices

Before SSH Operations

1. Check existing sessions

   ssh_session_list
   

2. Check recent workdir

   ls ~/.ssh-workdir/{hostname}/
   

3. Create new workdir if starting new task

   mkdir -p ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/output
   

During SSH Operations

1. Use appropriate tool for the task

   • Single command: ssh_execute
   • Multi-step: ssh_session_start → ssh_session_send → ssh_session_close
   • File transfer: ssh_upload/download/sync
   • Monitoring: ssh_monitor, ssh_tail, ssh_health_check

2. Log commands to workdir

   echo "command" >> ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/commands.md
   

3. Save important outputs

   ssh_execute server="{hostname}" command="df -h" > ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/output/df-h.txt
   

After SSH Operations

1. Close sessions

   ssh_session_close session="{session-id}"
   

2. Write summary

   echo '## Findings' >> ~/.ssh-workdir/{hostname}/{YYYY-MM-DD}-{topic}/summary.md
   

3. Clean up

   • Close tunnels: ssh_tunnel_close
   • Verify all sessions closed: ssh_session_list

Tips

• Use ssh_session_list before starting new tasks to reuse existing sessions
• Create workdir for each task to maintain organized history
• Write summaries to quickly recall previous work
• Use ssh_connection_status action="status" to check connection health
• For server comparison, store outputs with consistent naming across hosts
• Close sessions when done to free resources
• Configure server-side ClientAliveInterval for longer session timeouts if needed

Additional Resources

Reference Files

• references/sessions.md - Session management deep dive
• references/workspace.md - Workdir structure and usage
• references/comparison.md - How to compare historical data

Example Files

• examples/system-check.md - Complete system health check workflow
• examples/deployment.md - Multi-step deployment example
• examples/troubleshooting.md - Problem diagnosis workflow

Scripts

• scripts/create-workdir.sh - Create new workdir structure
• scripts/log-command.sh - Log command to workdir
• scripts/save-status.sh - Capture and save host status