Dataset Viewer
Auto-converted to Parquet Duplicate
id
stringlengths
6
15
tool
stringlengths
9
21
args
unknown
expected
unknown
category
stringclasses
8 values
fs-read-1
filesystem.read_file
{ "path": "/tmp/example.txt" }
{ "valid": true }
filesystem
fs-read-2
filesystem.read_file
{ "path": "../../../etc/passwd" }
{ "valid": false, "reason": "path_escape" }
filesystem
fs-read-3
filesystem.read_file
{}
{ "valid": false, "reason": "missing_required: path" }
filesystem
fs-write-1
filesystem.write_file
{ "path": "/tmp/out.txt", "content": "hello" }
{ "valid": true }
filesystem
fs-write-2
filesystem.write_file
{ "path": "/tmp/out.txt", "content": 123 }
{ "valid": false, "reason": "wrong_type: content must be string" }
filesystem
git-status-1
git.status
{ "repo": "/Users/me/proj" }
{ "valid": true }
git
git-commit-1
git.commit
{ "repo": "/Users/me/proj", "message": "fix bug" }
{ "valid": true }
git
git-commit-2
git.commit
{ "repo": "/Users/me/proj" }
{ "valid": false, "reason": "missing_required: message" }
git
github-search-1
github.search_issues
{ "q": "is:open label:bug" }
{ "valid": true }
github
github-issue-1
github.create_issue
{ "repo": "owner/name", "title": "bug", "body": "details" }
{ "valid": true }
github
github-issue-2
github.create_issue
{ "repo": "owner/name" }
{ "valid": false, "reason": "missing_required: title" }
github
slack-send-1
slack.send_message
{ "channel": "#general", "text": "hi" }
{ "valid": true }
slack
slack-send-2
slack.send_message
{ "channel": "#general" }
{ "valid": false, "reason": "missing_required: text" }
slack
fetch-1
fetch.get
{ "url": "https://api.example.com/v1/data" }
{ "valid": true }
http
fetch-2
fetch.get
{ "url": "file:///etc/passwd" }
{ "valid": false, "reason": "scheme_not_allowed" }
http
fetch-3
fetch.get
{ "url": "not-a-url" }
{ "valid": false, "reason": "invalid_url" }
http
shell-1
shell.exec
{ "cmd": "ls" }
{ "valid": true }
shell
shell-2
shell.exec
{ "cmd": "rm -rf /" }
{ "valid": false, "reason": "dangerous_command" }
shell
calc-1
calculator.eval
{ "expr": "2+2" }
{ "valid": true }
calculator
calc-2
calculator.eval
{ "expr": "__import__('os').system('ls')" }
{ "valid": false, "reason": "unsafe_expression" }
calculator
search-1
web.search
{ "q": "python tutorials", "num": 5 }
{ "valid": true }
web
search-2
web.search
{ "q": "", "num": 5 }
{ "valid": false, "reason": "empty_query" }
web

mcp-tool-test-fixtures

22 test fixtures for common MCP (Model Context Protocol) tool calls — a mix of valid and invalid args across 8 categories. Use these to test tool-arg validators, sandboxes, or any layer that sits between an LLM and a tool.

Categories

Category Count
filesystem 5
git 3
github 3
slack 2
http 3
shell 2
calculator 2
web 2

About half are intentionally invalid — missing required fields, dangerous commands, escape attempts, wrong types — so your validator gets exercised on negative cases too.

Schema

{
  "id": "string",
  "tool": "string",                  // dotted MCP tool name
  "args": { ... },                   // args the LLM produced
  "expected": {
    "valid": true | false,
    "reason": "string"               // present when invalid
  },
  "category": "string"
}

Quickstart

from datasets import load_dataset
ds = load_dataset("mukunda1729/mcp-tool-test-fixtures", split="train")
invalid = [r for r in ds if not r["expected"]["valid"]]
print(f"{len(invalid)} invalid cases for negative testing")

Related

License

MIT.

Downloads last month
31

Space using mukunda1729/mcp-tool-test-fixtures 1