scope

Manages per-channel opt-in access scope, trust files, and execution gating for messaging plugins.

62|14|Updated Apr 9, 2026
One-click install
npx skills add https://github.com/crisandrews/ClawCode --skill scope-crisandrews
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: scope
Source: https://github.com/crisandrews/ClawCode/tree/main/skills/scope
Command: npx skills add https://github.com/crisandrews/ClawCode --skill scope-crisandrews

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When ClawCode connects to messaging plugins like claude-whatsapp, memory tools could expose chat content to non-owner senders. This Skill configures and audits the channel-scope layer that filters memory access per channel and per chat, so shared channels cannot leak private conversations. ## Core Features & Use Cases - Status and audit: Inspect each channel's scope mode, identity, execGate policy, and trust file presence, plus doctor rows for scope health. - Enable, disable, and wizard flows: Turn shadow or enforce mode on per channel via validated Bash writes, or walk through an interactive wizard covering identity, background lane, and execution gating. - Dry-run testing: Probe whether a given chat ID or sender-plus-tool combination would be allowed under the current scope before enforcing it. - Use Case: After pairing claude-whatsapp, run the wizard to enable shadow mode, create the per-workspace owner trust file, and preview how many channel chunks would be filtered for non-owner queries. ## Quick Start Ask the agent to run the scope status command to show the current channel scope configuration and trust file state.

Frequently Asked Questions about scope

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

FAQPage Schema
How do I enable channel scope for WhatsApp in ClawCode?

Run the scope wizard for an interactive setup, or use the enable subcommand with the channel and mode, such as enabling whatsapp in shadow or enforce mode. Scope writes go through Bash because the config tool refuses scope keys, so you approve each change via a permission prompt.

What is the difference between shadow and enforce mode in channel scope?

Shadow mode collects statistics on what would be filtered without actually blocking anything, which is recommended for the first week. Enforce mode actively filters denied chunks so non-owner queries cannot see restricted channel content.

Why does owner identity in config not unlock full memory access?

Owner unlock requires two factors: the identity setting in config plus an out-of-band per-workspace trust file created via Bash. Config alone never unlocks, which prevents a prompt-injected agent from declaring itself owner by writing config.

Does channel scope block Read or Grep on channel log files?

No. Scope filtering applies only at the MCP tool level for memory and messaging tools. Native Read, Grep, and direct SQLite access bypass scope by design; hard isolation requires OS-level filesystem permissions.

How do I test if a chat would be visible before enforcing scope?

Use the test subcommand with a chat ID to run a dry-run probe against the scope adapter. It reports whether the chat would be visible under the current mode without changing any config or memory state.

What does the execution gate block for non-owner messages?

The execGate blocks destructive tools like Bash, Write, Edit, and agent_config when a turn was triggered by a non-owner inbound message. It supports denylist or allowlist policies, and a separate per-workspace exec trust file can unlock it.