tool-abuse

tool-abuse is a cursor rule for Cursor from davidmatousek/tachi. It costs 40 tokens per session (2,293 once invoked), scanned A, original, Apache-2.0.

A security rule set for checking how AI agents use external tools such as plugins, MCP servers, and function-calling APIs. It looks for unsafe tool access, altered parameters, dangerous tool combinations, and malicious tool definitions.

In plain words
What is it for?
It is for reviewing AI-agent workflows that invoke tools, servers, plugins, or APIs. It can identify unauthorized calls, privilege escalation, parameter injection, unsafe call chains, and tool poisoning.
Why use it?
It helps reveal ways an agent could gain abilities it should not have or be manipulated into taking unsafe actions. This reduces risks caused by trusting tools and tool instructions without enough checks.

Cursor rule for Cursor

Written for Cursor: a Cursor rule (.mdc).

Needs its repository: it reads a path above its own folder, which exists only inside the repository. The line is output_schema: ../../../schemas/finding.yaml.

Good fit It is for reviewing AI-agent workflows that invoke tools, servers, plugins, or APIs. It can identify unauthorized calls, privilege escalation, parameter injection, unsafe call chains, and tool poisoning.

Compare 6 cursor rules from other repositories ↓
Install

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.

Clone the repo
git clone --depth 1 https://github.com/davidmatousek/tachi
agentmods
npx agentmods add rules/davidmatousek/tachi/tool-abuse

Made for: Cursor.

Wrote 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.

agentmods badge for tool-abuse

README.md
[![agentmods](https://agentmods.dev/badge/rules/davidmatousek/tachi/tool-abuse.svg)](https://agentmods.dev/rules/davidmatousek/tachi/tool-abuse)
Your own site
<a href="https://agentmods.dev/rules/davidmatousek/tachi/tool-abuse"><img src="https://agentmods.dev/badge/rules/davidmatousek/tachi/tool-abuse.svg" alt="Measured on agentmods" height="20"></a>
Per session 40 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,293 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00040 $0.02293
Opus 5 $0.00020 $0.01146
Sonnet 5 $0.00008 $0.00459
Haiku 4.5 $0.00004 $0.00229

Measured 3d ago against content hash b69a3b672535, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

tool-abuse scanned grade A with 0 findings 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 3d 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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

adapters/cursor/rules/tool-abuse.mdc · 182 lines

How it starts

The opening of the file, as written. The whole thing — 182 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Metadata

category: agentic
threat_class: AG
dfd_targets: [Process]
owasp_references: [ASI-02, ASI-04, MCP-03, MCP-05]
output_schema: ../../../schemas/finding.yaml

Tool-Use Abuse Threat Agent

Purpose

Detects threats arising from agentic systems that invoke external tools, MCP servers, plugins, or APIs as part of their execution. Tool-use abuse occurs when an agent invokes tools it should not have access to, manipulates tool parameters to achieve unintended effects, or chains tool calls in sequences that escalate privileges beyond the agent's intended scope. This agent identifies unauthorized tool invocation, capability escalation through tool composition, parameter injection in tool calls, tool chain manipulation where individually safe operations combine to produce dangerous outcomes, and tool poisoning attacks (direct poisoning, tool shadowing, and rug pulls) where malicious tool definitions alter agent behavior.

Detection Scope

Trigger Keywords

This agent activates when a DFD element name or description matches any of the following patterns (case-insensitive):

  • agent
  • MCP server
  • tool server
  • plugin
  • tool use
  • function calling
  • tool chain
  • orchestrator
  • action executor

Applicable DFD Element Types

  • Process: Any process node that represents an agent, orchestrator, or tool execution layer. This includes MCP client processes, plugin hosts, function-calling middleware, and agentic frameworks that dispatch tool invocations based on model output.

Empty Results Guidance

If the architecture input contains no components matching the trigger keywords above (no agents, MCP servers, tool servers, plugins, function-calling layers, or tool chains), this agent MUST produce zero findings. Do not generate speculative findings about hypothetical tool-use components. An architecture without tool servers, MCP integrations, plugin systems, or agentic tool invocation is outside this agent's detection scope. Return an empty findings list.

Read the full file on GitHub · 182 lines

Changes

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.

  1. 3d ago First seen · 182 lines · 40 tokens per session scan A b69a3b672535

Subscribe to this mod's changes

tool-abuse is a cursor rule published in the GitHub repository davidmatousek/tachi (90 stars, last pushed 25d ago), licensed Apache-2.0. It adds 40 tokens to every session and 2,293 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.