A dynamic browsing assistant that transforms high-level goals into actionable checklists and persists session data for structured analysis.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install goal-mode
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 goal-mode 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 Goal Mode?
Goal Mode is a sophisticated browsing analysis engine designed for AI agents within the Openclaw Skills ecosystem. It acts as a goal-aware orchestrator that receives structured JSON requests to manage complex tasks like research, shopping, or learning. By generating specific evaluation criteria and maintaining a persistent workspace, it ensures that every browsing session is methodical, documented, and results-driven.
This skill is built for reliability, enforcing a strict execution model where all workspace writes and data persistence must occur before a final response is delivered. It bridges the gap between raw web browsing and structured knowledge management, making it an essential component for any agent requiring long-term session awareness and objective-based navigation.
Goal Mode Use Cases
- Researching complex technical topics with automated criteria generation.
- Comparing products or services through structured evaluation of multiple web pages.
- Planning travel or events by tracking progress against specific requirements.
- Career exploration and deep-dive learning sessions with automated session wrap-ups.
- Decision-making tasks that require evidence-based recommendations and confidence scoring.
How Goal Mode Works
- The skill receives a structured JSON request specifying an operation such as generate_criteria, evaluate_page, or resume_goal.
- It validates the input and reads internal reference files along with shared schemas to ensure data consistency.
- The engine computes the analysis (e.g., scoring page relevance against criteria) while maintaining the state in memory.
- It executes mandatory workspace persistence by writing updated session data, event logs, and human-readable summaries to the local filesystem.
- Once all tool calls and writes are complete, it outputs a single raw JSON response containing the final analysis and status.
Goal Mode Setup
To utilize this skill within the Openclaw Skills framework, ensure your environment has the necessary directory permissions for the workspace root. The skill automatically manages its own file structure.
# Ensure the workspace directory exists
mkdir -p /home/ubuntu/.openclaw/workspace/goal-mode
# The skill requires the following reference files in your baseDir:
# - references/generate-criteria.md
# - references/evaluate-page.md
# - references/schemas.md
Goal Mode Data Schema & Taxonomy
Goal Mode maintains a highly organized directory structure for session tracking and long-term memory. Data is stored in /home/ubuntu/.openclaw/workspace/.
| Path | Purpose |
|---|---|
goal-mode/active-goal.json |
Tracks the current active session slug. |
{goal_slug}/session.json |
Stores the complete state including findings and criteria coverage. |
{goal_slug}/events/ |
Contains immutable, timestamped logs of every page evaluation. |
memory/goal-mode/active-session.md |
Provides a human-readable live status of the current goal. |
memory/goal-mode/history.md |
An append-only log of all past session outcomes. |
Goal Mode Advanced Features
- Intelligent criteria specificity: Generates narrow, actionable criteria to prevent broad, low-value analysis.
- Depth-over-breadth scoring: Implements a relevance threshold (0.7+) ensuring that only detailed coverage satisfies goal criteria.
- Per-page coverage caps: Limits a single page to satisfying a maximum of 3 criteria to encourage diverse source exploration.
- Context-aware findings: Dynamically adjusts finding types based on the goal (e.g., technical specs for research vs. pricing for shopping).
- Automated session resume: Allows agents to pick up exactly where they left off by reading the persisted workspace state.
SKILL.md
Loading
Related Openclaw Skills
WHOOP Fitness Tracker Integration
A comprehensive integration to sync WHOOP biometric data with AI agents for automated recovery reporting and physiological-aware productivity coaching.
SEO Audit
A comprehensive diagnostic tool that audits webpage HTML for SEO compliance and generates prioritized optimization reports within Openclaw Skills.
Invoice Generator
A specialized skill for creating professional, print-ready HTML invoices directly from natural language or structured data inputs.
Landing Page Builder
An automated web generation tool that transforms business ideas into polished, single-page landing pages using structured templates or natural language prompts.
Alpaca Trading Skill
A programmatic interface for executing stock and crypto trades through the Alpaca API using Openclaw Skills.
Proactive Agent
A skill that transforms passive AI agents into anticipatory assistants that take initiative, predict user needs, and provide proactive status updates.
Featured*
