More screenshots
Works with the AI Agents you already use
FileReadWrite
Manage files and run gated commands on local or remote machines via the FileReadWrite MCP server.
Free
FileReadWrite
Example session with this skill installed
List my connected machines, find the 'logs' folder in the allowed roots of the first machine, and read the last 50 lines of 'error.log' inside it.
- Read your context and instructions
- Compiled the filereadwrite
I have identified your active machines and located the requested log file. I used getAgents to find the active ID, listAllowedResources to confirm the path, and readFile with a line range to avoid truncation.
Remote File Access Summary
| Action | Details | Result |
|---|---|---|
| Target Machine | mbp-work-01 (ID: ag_7f29) | Connected |
| Path Resolved | /Users/dev/project/logs/error.log | Validated |
| Read Operation | Lines 450-500 | Success |
| File Hash | sha256:8f31b2... | Stored for edits |
Log Preview (error.log)
[2023-10-27 14:22:01] ERROR: Connection timeout in module 'auth'
[2023-10-27 14:22:05] WARN: Retrying connection (1/3)
[2023-10-27 14:22:10] ERROR: Fatal exception: Null pointer in 'session_manager.py' at line 122
...
[2023-10-27 14:23:45] INFO: System heartbeat stable.
Next steps
- If you want to fix the exception in
session_manager.py, I can read that file next. - I can run a grep search across the entire
logsfolder for "Fatal exception" to see how often this occurs. - If you have the
pytestexecutable in your allow-list, I can trigger a test run to verify the fix.
Connects securely to your tools. The creator never sees your data.
About this skill
Managing files across multiple remote machines is often blocked by high-latency SSH sessions or complex permissions. This skill provides a structured interface for the FileReadWrite MCP server to interact with sandboxed local environments.
What it does
- Multi-machine discovery: Lists online agents and targets specific machines via unique IDs.
- Sandboxed exploration: Maps allowed root directories and executables to ensure operations stay within user-defined boundaries.
- Robust file reading: Handles large files with line-range pagination and detects truncation automatically.
- Safe write protocol: Uses a hash-based verification system to prevent race conditions or accidental overwrites during concurrent edits.
- Controlled command execution: Triggers allowed binaries through an approval-gated runner that requires manual user confirmation.
- Deep search: Executes recursive regex searches across filenames or file contents while respecting common ignore patterns like
node_modules.
How it works
- Call
getAgentsto identify online machines and select a targetagentId. - Use
listAllowedResourcesto verify the active allow-list for paths and executables. - Read target files to obtain the current content and a mandatory integrity hash.
- Submit edits using
writeFilewith the matching hash to update specific line ranges or entire files. - Invoke
runCommandfor build or test tasks, then wait for the user to approve the execution on their local terminal.
Frameworks & tools
This skill requires an active connection to a FileReadWrite MCP server and the frw-agent binary running on the target machine.
Why this beats prompting it yourself
Standard LLM prompts often fail when files are too large or when file system states change between turns. This skill enforces a hash-check protocol that prevents data corruption and provides specific error handling for common local-environment failures.
Use cases
- Debugging logs on a remote staging server by reading specific line ranges.
- Refactoring code across multiple machines while ensuring file integrity.
- Running allowed test suites or build scripts with manual approval gates.
- Searching for specific regex patterns across a large local codebase.
Known limitations
Does not support binary file writes or shell-specific syntax like pipes and redirects. Operations outside user-defined allow-lists will return access denied errors.
Requires the FileReadWrite MCP server. No support for binary data eg images, videos, pdfs or redirects. Commands time out after 120 seconds. Commands require manual user approval.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 20 days ago
- Free with an Agensi account
Requires
Permissions
Allowed Hosts