agent-tools

Configure Claude Code agent tool permissions with path and command restrictions.

14|1|Updated Dec 5, 2025

Search Tags:#security #cli #permissions #claude-code #tools #restrictions

Ready to run

npx skills add https://github.com/jpoutrin/product-forge --skill agent-tools

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: agent-tools
Source: https://github.com/jpoutrin/product-forge/tree/main/plugins/product-design/skills/agent-tools
Command: npx skills add https://github.com/jpoutrin/product-forge --skill agent-tools

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill provides a centralized reference for configuring tool permissions when launching Claude Code agents, helping reduce risk and misconfigurations.

Core Features & Use Cases

List of available tools with descriptions
How to set strict allowed-tools and path restrictions
Guidance on per-task or per-agent constraints

Quick Start

Use the --allowedTools flag to restrict access, for example: claude --allowedTools "Read" "Grep" "WebFetch" --print "run analysis"

Frequently Asked Questions about agent-tools

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I restrict tool permissions for Claude Code agents?▼

Restrict tool permissions using the --allowedTools flag when launching Claude Code agents. Specify which tools agents can access by name, such as --allowedTools "Read" "Grep" "WebFetch", to prevent unauthorized access and reduce security risk during deployment.

What tools can I restrict access to in Claude Code?▼

Claude Code supports restricting access to individual tools, file paths, Bash commands, WebFetch domains, and MCP tools. The agent-tools Skill provides a centralized reference listing available tools with descriptions and guidance on configuring granular per-task or per-agent constraints.

Can I set different tool permissions for different agents?▼

Yes, tool permissions can be configured per-task or per-agent during Claude Code deployment. Use the --allowedTools flag and path-based restrictions to apply granular constraints tailored to each agent's specific security requirements and functional needs.

How do I prevent agents from accessing certain files or directories?▼

Configure path-based restrictions through tool permission settings to limit file access. The agent-tools Skill provides example configurations for restricting file read and write operations, supporting both parallel and granular deployment scenarios.

What are the security risks of unrestricted Claude Code tool access?▼

Unrestricted tool access increases risk of unauthorized file access, unintended command execution, and uncontrolled external requests. Defining strict allowed-tools and implementing path restrictions mitigates these risks during agent deployment and reduces misconfigurations.

Can I restrict Bash commands and WebFetch domains separately?▼

Yes, tool permissions support independent configuration of Bash command restrictions and WebFetch domain controls. The agent-tools Skill provides separate guidance on each constraint type, enabling fine-grained security policies for different agent workflows.