allow-agent-commands

name	allow-agent-commands
description	Add or change allowed commands in AI agent permission configs (OpenCode, Claude Code)

Adding/Changing Allowed Commands

When to use this skill

Use this skill when:

• A command needs to be added to the allow list for AI agents
• Permission settings need to be modified for existing commands
• Setting up a new AI tool that requires command permissions

How it works

Commands must be explicitly allowed in the permission configs before agents can run them. This skill covers adding entries to both OpenCode and Claude Code configuration files.

IMPORTANT: The config files are stowed into ~ via GNU Stow symlinks. Always edit them directly from within the ai-config repository, NOT from ~. The paths below are relative to the ai-config repository root.

OpenCode Configuration

File: .config/opencode/opencode.json under permission.bash

Format:

"uname *": "allow",

Pattern: "command argpatterns": "action" where * matches any args

Important: Insert new entries in alphabetical order by command name. When adding a new command, find the correct position to maintain sorting. Do NOT reorder existing entries.

Claude Code Configuration

File: .claude/settings.json under permissions.allow

Format:

"Bash(uname:*)",

Important: Insert new entries in alphabetical order. Commands are sorted by the command name (the part after Bash( and before :). Do NOT reorder existing entries.

General Pattern for New AI Tools

When adding new AI tools:

1. Find their config file (usually in ~/.config/ or project root)
2. Look for permissions, allowedCommands, or similar sections
3. Add the command with glob patterns for arguments
4. Insert new entries in alphabetical order by command name
5. Do NOT reorder existing entries.

After Adding Commands

After adding the new command entries to both config files:

1. Check if there are any other staged changes with git status
2. If there are NO other staged changes, stage ONLY the new hunks in the two config files using the git-stager subagent (be careful not to stage unrelated changes)
3. Invoke the git-committer subagent to commit the changes

Examples

Adding rg (ripgrep) to both configs:

OpenCode (insert after rev *, before rpm -q*):

"rg *": "allow",

Claude Code (insert after Bash(rg:*), before Bash(rpm -q:*)):

"Bash(rg:*)",