Grit is a relentless problem-solving skill that enables AI agents to escalate tools, iterate on failures, and push through technical blockers until a task is completed.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install grit
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 grit 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 Grit?
Grit transforms an AI agent from a passive follower into a stubborn, practical fixer. This skill is designed for scenarios where standard automation fails due to flaky UIs, brittle locators, or environmental constraints. By utilizing Grit within Openclaw Skills, an agent stops viewing failure as an end-point and instead treats it as data for the next logical attempt.
The skill operates on a philosophy of methodical escalation. It empowers the agent to switch interaction modes—such as moving from DOM-based clicking to coordinate-based clicking—or even researching and installing entirely new software to get the job done. While it increases persistence, it remains strictly subordinate to workspace security protocols and system safety rules.
Grit Use Cases
- Overcoming brittle browser automation when standard Playwright or DOM locators fail to interact with complex web elements.
- Managing end-to-end deployment workflows where signing, versioning, or platform-specific UI hurdles frequently block progress.
- Automating tasks on platforms with aggressive anti-bot measures or inconsistent interface states.
- Situations requiring the research, installation, and immediate deployment of new specialized tools to solve a specific bottleneck.
- Tasks where the user provides a do whatever it takes mandate to ensure completion without constant human intervention.
How Grit Works
- Blocker Definition: The agent identifies and states the exact technical failure in a single, concrete sentence to focus the recovery effort.
- Direct Correction: A focused attempt is made using the current toolchain to rule out simple state-based errors.
- Tactical Switching: If the direct fix fails, the agent rotates interaction methods, such as moving from API calls to CDP event injection or GUI automation.
- Tool Escalation: The agent researches and installs purpose-built Openclaw Skills or CLI tools that are better suited for the specific obstacle.
- Security Verification: Before using any newly installed tool, the agent runs mandatory SOP security scans and audits to maintain environment integrity.
- Iterative Refinement: The agent loops through attempts, ensuring each cycle changes a material variable like the tool, target state, or sequencing to avoid blind retrying.
- Final Resolution: The process continues until the task is successful or an immovable external blocker, such as a 2FA requirement, is identified.
Grit Setup
To integrate Grit into your agent environment, ensure it is available in your workspace and accessible via standard trigger phrases. You can initialize the environment or check for required SOP files using the following commands:
# Check for existing SOP or workspace rules
cat SOP.md
# Trigger grit-based problem solving via CLI if supported
openclaw-agent "use the grit skill to fix the deployment pipeline"
Ensure that an escalation-ladder.md is present in your references/ directory to provide the agent with a structured hierarchy for tool selection.
Grit Data Schema & Taxonomy
Grit organizes its operational logic and progress tracking through a structured approach to workspace files and session memory.
| Data Type | Description | Persistence |
|---|---|---|
| SOP.md | The primary operating manual governing tool choice, security scans, and logging. | Workspace Root |
| escalation-ladder.md | A reference document defining the hierarchy of tools from least to most invasive. | references/ |
| Task State | Internal tracking of solved vs. unsolved layers to prevent regression during retries. | Session Memory |
| Audit Logs | Records of tool installations and security scan results performed during escalation. | Workspace Logs |
| Screenshots | Visual data used for diagnosis when switching from DOM to GUI-based automation. | temp/ |
Grit Advanced Features
- Auto-Escalation: Automatically moves from high-level APIs to low-level event injection without manual prompts.
- Dynamic Skill Acquisition: Capable of searching for, auditing, and installing new Openclaw Skills mid-workflow to bridge capability gaps.
- SOP Adherence: Strictly follows local workspace security protocols, ensuring that increased persistence does not bypass safety guardrails.
- Isolation Logic: Separates successfully completed sub-tasks from currently blocked ones to ensure the agent never loses progress during complex iterations.
SKILL.md
Loading
Related Openclaw Skills
WeChat Official Accounts API
A specialized skill for extracting and analyzing WeChat Official Accounts data including engagement metrics, comments, and published posts.
Openclaw Engineering Harness
A professional-grade engineering loop for AI agents that ensures request clarification, minimal code changes, and validated delivery.
YouTube Channel Compare
A professional benchmarking tool to compare performance metrics and engagement across multiple YouTube channels.
Xiaohongshu Creator Marketplace (Pugongying) API
A specialized integration for accessing Xiaohongshu Pugongying creator metrics, profiles, and historical growth trends.
Feishu CLI Setup
A specialized AI agent guide for the seamless installation, configuration, and OAuth authentication of the official Lark and Feishu CLI.
Memory Auditor
Memory Auditor is an automated service that detects context drift and memory fabrication by comparing agent claims against stored logs.
Featured*
