- Home
- Skills
- Testing & Debugging
- Windows PowerShell Safe
Works with the AI Agents you already use
Windows PowerShell Safe
Help AI agents avoid common Windows PowerShell quoting, encoding, version and exit-code mistakes.
Free
Windows PowerShell Safe
Example session with this skill installed
On Windows PowerShell 5.1, demonstrate how to pass a path and label containing spaces to a native executable and capture a nonzero exit code. Use a harmless fixture; do not read or modify any real project data.
- Read your context and instructions
- Compiled the windows powershell safe
Checked behavior
The native fixture received exactly four argument items. The path containing a space and the multiword label were each preserved as one item. The fixture deliberately exited with code 7, and PowerShell captured that code immediately.
Invocation pattern
$argList = @($fixture, '--input', 'C:\Demo Data\input.txt', '--label', 'example label')
& $Python @argList
$code = $LASTEXITCODE
if ($code -ne 7) { throw "Expected fixture exit 7, got $code" }
Here, $Python is a verified absolute interpreter path and $fixture is the harmless demo-argv.py file supplied with the submission materials. C:\Demo Data\input.txt is only an argument string: the fixture does not open it.
Observed output
{"arguments": ["--input", "C:\\Demo Data\\input.txt", "--label", "example label"], "argument_count": 4}
Native exit code: 7
PASS: argument boundaries and nonzero native exit code verified.
Next steps and scope
Use the same argument-array and immediate-exit-code pattern with the project's approved executable. Handle the actual tool's success codes according to its documentation. This check covers Windows PowerShell 5.1 invocation only; it does not prove that every command, encoding case or PowerShell 7 environment is correct.
Connects securely to your tools. The creator never sees your data.
About this skill
A free, focused guide for AI coding agents that run commands on Windows. It covers problems that frequently break otherwise reasonable automation: mixing Windows PowerShell 5.1 with PowerShell 7, nested quoting, paths with spaces, text encoding, native exit codes and process launching.
What it teaches
- Distinguish powershell.exe from pwsh.exe before choosing syntax or .NET APIs.
- Pass native arguments as separate array items instead of composing a shell string.
- Prefer a .ps1 file over fragile nested -Command expressions.
- Treat native exit codes and PowerShell cmdlet exceptions separately.
- Preserve existing file encodings and explicitly choose output encodings.
- Stop repeated failures and validate destructive paths before proceeding.
What's included
SKILL.md, eight experience/documentation-based failure examples, source metadata and both upstream MIT notices. The main instructions are in Simplified Chinese; code and command identifiers remain unchanged. The marketplace description is English.
Evidence and limits
A harmless Windows PowerShell 5.1 demonstration verified argument boundaries for paths/labels with spaces and immediate capture of a native process's deliberate exit code 7. This is a focused invocation check, not a complete agent-behavior or PowerShell 7 regression suite. The eight reference examples remain labeled as experience/documentation guidance rather than fully rerun tests. This skill is not an OS sandbox and cannot guarantee safe execution.
Attribution
Adapted and maintained by haxsd / agent-skills contributors from Misaka Mikoto's powershell-safe-invocation (MIT), with stop-rule wording adapted from agent-shells contributors (MIT). Original copyright notices and fixed source revisions are included. Original contribution is the local adaptation, not ownership of the upstream method. No upstream endorsement is claimed.
No API key or paid service is required to use the guide. Reading it does not require administrator privileges. File changes and process execution still require authorization for the underlying task.
ProcessStartInfo.ArgumentList is unavailable on PowerShell 5.1. Requires manual scrubbing of private logs. Destructive operations are blocked on root-level or empty paths.
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
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
Permissions
Terminal/process execution and local file read/write only when needed for the user-authorized task. The guide itself makes no network requests and requires no credentials. No administrator privileges by default. This is guidance, not an OS sandbox.