Scope catches AI agent actions that cross the boundaries agreed during reconnaissance and pauses for your decision.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install scope
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install scope using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Scope?
Scope is a drift guard for AI coding agents: it compares a pending file change, shell command, dependency addition, or refactor against the original task and the boundaries recorded during recon. When it detects genuine scope creep, it stops before acting and produces a specific report showing what boundary would be crossed.
As an Openclaw Skills tool, Scope helps keep small development tasks small without interrupting normal in-scope work. You choose whether the agent may proceed with a justification, should defer the work to a follow-up, or should revert to the last in-scope state.
Scope Use Cases
- Catch an agent preparing to edit a file that recon did not identify as in scope.
- Pause before a shell command would affect resources outside the project directory.
- Review unplanned dependency additions or refactors unrelated to the task.
- Stop actions justified only by reasoning such as “while I'm here.”
- Keep the user in control when an agent encounters a legitimate need to expand the task.
- Distinguish real scope drift from incomplete recon, required test files, and trivial formatting changes to in-scope files.
How Scope Works
- The agent establishes the original task and in-scope files or resources during recon.
- Before a potentially out-of-scope action, Scope compares the action with the task description and recon's proposed boundaries.
- If the action is genuine drift, Scope pauses before execution and displays a structured report identifying the action, justification, affected resources, and precise boundary mismatch.
- The report offers three choices: proceed with justification, defer the work to a follow-up, or revert the change and return to the last in-scope state.
- Scope waits for the user's choice and takes no further action until the user responds.
- After the decision, it follows the selected path: log the justification and proceed, add the item to the follow-up list and resume in-scope work, or revert and resume from the last in-scope state.
Scope Setup
The provided SKILL.md defines Scope's behavior but does not specify an installer, CLI command, configuration file, or required dependencies. Add the skill to the skills directory or registry used by your AI agent, following that runtime's instructions, and ensure the agent can access the task description and recon findings.
# No installation command is specified in the provided SKILL.md.
# Install or register the skill using your agent runtime's documented process.
For reliable drift checks, record the task summary and the files or resources found during recon. No additional setup options are documented.
Scope Data Schema & Taxonomy
Scope uses the task and recon findings as its comparison baseline. The skill does not specify a persistent storage format, generated files, or metadata taxonomy.
| Information | Role |
|---|---|
| Original task summary | Defines the task contract used to assess a pending action. |
| Recon “What I found” section | Identifies files and resources already considered in scope. |
| Pending action | The file, resource, command, dependency, or refactor being evaluated. |
| Drift report | Markdown report with the original task, proposed action, scope mismatch, options, and decision prompt. |
| Session log | Receives the justification when the user chooses to proceed. |
| Follow-up list | Receives deferred work when the user chooses to defer. |
| Last in-scope state | Reference point for resuming after a revert. |
The report requires the agent to name the specific boundary being crossed; a generic out-of-scope warning is insufficient.
Scope Advanced Features
- Intercepts scope expansion before the proposed action is executed.
- Checks multiple drift signals, including unlisted file edits, commands affecting resources outside the project directory, unplanned dependencies, unrelated refactors, and “while I'm here” work.
- Uses a structured Markdown drift report so the user can assess the action and its risks.
- Offers explicit proceed, defer, and revert paths, then waits for user confirmation.
- Appends a justification to the session log when the user approves proceeding.
- Treats deferral as a valid default and distinguishes real drift from recon gaps, required tests, and trivial formatting changes.
- The provided skill does not specify hooks, multi-agent coordination, or automatic file logging beyond the described session-log and follow-up behavior.
SKILL.md
Loading
Related Openclaw Skills
Shell Chat
Shell Chat lets AI agents exchange real-time direct messages, group chats, photos, and files with people and other agents.
TikTok Comment Consumer Insights
Retrieve public TikTok search results, creator posts, and video comments as structured JSON to support consumer insights and content research.
Blave Quant
Blave Quant connects AI agents to global market data and exchange workflows, with explicit confirmation required for every trading action.
Chelaile Real-Time Bus
Query real-time bus arrivals, vehicle locations, routes, and public transit plans across 480 cities through Chelaile MCP.
Hold
Hold stops an AI agent after an unanticipated failure, reports what happened, and waits for the user to choose how to proceed.
Recon
Recon inspects a repository, surfaces risks and unknowns, and requires user confirmation of task-specific rules before coding begins.
Featured*
