ToolUniverse is a collection of tools, interfaces, and supporting components for building AI systems that perform scientific work. It is for developers creating AI scientist agents that use APIs, databases, machine-learning tools, and domain-specific utilities. The catalogue includes skills, commands, an MCP server, an agent, and a hook for working with the ecosystem.
Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/mims-harvard/ToolUniversenpx agentmods add skills/mims-harvard/tooluniverse/devtu-create-toolWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/mims-harvard/tooluniverse/devtu-create-tool)<a href="https://agentmods.dev/skills/mims-harvard/tooluniverse/devtu-create-tool"><img src="https://agentmods.dev/badge/skills/mims-harvard/tooluniverse/devtu-create-tool/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/mims-harvard/tooluniverse/devtu-create-tool"><img src="https://agentmods.dev/badge/skills/mims-harvard/tooluniverse/devtu-create-tool.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- Socket pass
- Snyk warn
- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Data Exfiltration · line 65 Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.Fix: Verify the destination URL is trusted and necessary. Remove or replace with documented APIs. Ensure no secrets, tokens, or PII are transmitted.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00075 | $0.01863 |
| Opus 5 | $0.00037 | $0.00932 |
| Sonnet 5 | $0.00015 | $0.00373 |
| Haiku 4.5 | $0.00007 | $0.00186 |
Grade A, and why
devtu-create-tool scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
response = requests.get( How it starts
The opening of the file, as written. The whole thing — 222 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ToolUniverse Tool Creator
Create new scientific tools following established patterns.
Top 7 Mistakes (90% of Failures)
- Missing
default_config.pyEntry — tools silently won't load - Non-nullable Mutually Exclusive Parameters — validation errors (#1 issue in 2026)
- Fake test_examples — tests fail, agents get bad examples
- Single-level Testing — misses registration bugs
- Skipping
test_new_tools.py— misses schema/API issues - Tool Names > 55 chars — breaks MCP compatibility
- Raising Exceptions — should return error dicts instead
Two-Stage Architecture
Stage 1: Tool Class Stage 2: Wrappers (Auto-Generated)
@register_tool("MyTool") MyAPI_list_items()
class MyTool(BaseTool): MyAPI_search()
def run(arguments): MyAPI_get_details()
One class handles multiple operations. JSON defines individual wrappers. Need BOTH.
Three-Step Registration
Step 1: Class registration via @register_tool("MyAPITool")
Step 2 (MOST COMMONLY MISSED): Config registration in default_config.py:
TOOLS_CONFIGS = {
"my_category": os.path.join(current_dir, "data", "my_category_tools.json"),
}
Step 3: Automatic wrapper generation on tu.load_tools()
Implementation Guide
Files to Create
src/tooluniverse/my_api_tool.py— implementationsrc/tooluniverse/data/my_api_tools.json— tool definitionstests/tools/test_my_api_tool.py— tests
Python Tool Class (Multi-Operation Pattern)
from typing import Dict, Any
from tooluniverse.tool import BaseTool
from tooluniverse.tool_utils import register_tool
import requests
@register_tool("MyAPITool")
class MyAPITool(BaseTool):
BASE_URL = "https://api.example.com/v1"
def __init__(self, tool_config):
super().__init__(tool_config)
self.parameter = tool_config.get("parameter", {})
self.required = self.parameter.get("required", [])
def run(self, arguments: Dict[str, Any]) -> Dict[str, Any]:
operation = arguments.get("operation")
if not operation:
return {"status": "error", "error": "Missing: operation"}
if operation == "search":
return self._search(arguments)
return {"status": "error", "error": f"Unknown: {operation}"}
def _search(self, arguments: Dict[str, Any]) -> Dict[str, Any]:
query = arguments.get("query")
if not query:
return {"status": "error", "error": "Missing: query"}
try:
response = requests.get(
f"{self.BASE_URL}/search",
params={"q": query}, timeout=30
)
response.raise_for_status()
data = response.json()
return {"status": "success", "data": data.get("results", [])}
except requests.exceptions.Timeout:
return {"status": "error", "error": "Timeout after 30s"}
except requests.exceptions.HTTPError as e:
return {"status": "error", "error": f"HTTP {e.response.status_code}"}
except Exception as e:
return {"status": "error", "error": str(e)}
What ships with it
10 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- references/advanced-patterns.md 30 KB
- references/implementation-guide.md 11 KB
- references/quick-reference.md 11 KB
- references/testing-guide.md 2.8 KB
- references/tool-improvement-checklist.md 21 KB
- templates/api_tool_template.py 5.3 KB runs code
- templates/api_tools_config.json 5.6 KB
- templates/simple_tool_template.py 2.8 KB runs code
- templates/simple_tools_config.json 1.7 KB
- templates/test_template.py 7.7 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago First seen · 222 lines · 75 tokens per session scan A fbbcb3a10032
devtu-create-tool is a skill published in the GitHub repository mims-harvard/ToolUniverse (1,680 stars, last pushed 2d ago), licensed Apache-2.0. It adds 75 tokens to every session and 1,863 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
healthcare-fhir
Design RESTful clinical data exchanges using HL7 FHIR standards.
cqrs-implementation
Implement Command Query Responsibility Segregation for scalable architectures. Use when separating read and write models, optimizing query performance, or building event-sourced systems.
earth2studio-create-datasource
Create and validate Earth2Studio data source wrappers (DataSource, ForecastSource, DataFrameSource, ForecastFrameSource) from remote stores. Do NOT use for fetching data with existing sources, model inference, or installation tasks.
azure-communication-common-java
Azure Communication Services common utilities for Java. Use when working with CommunicationTokenCredential, user identifiers, token refresh, or shared authentication across ACS services.
spring-boot-event-driven-patterns
Provides Event-Driven Architecture (EDA) patterns for Spring Boot — creates domain events, configures ApplicationEvent and @TransactionalEventListener, sets up Kafka producers and consumers, and implements the transactional outbox pattern for reliable distributed messaging. Use when implementing event-driven systems…
nestjs-best-practices
Provides comprehensive NestJS best practices including modular architecture, dependency injection scoping, exception filters, DTO validation with class-validator, and Drizzle ORM integration. Use when designing NestJS modules, implementing providers, creating exception filters, validating DTOs, or integrating Drizzle…