A free, zero-dependency token consumption and cost analyzer that scans OpenClaw agent workspaces across 6 major language models.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install token-audit
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 token-audit 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 Token Audit?
The Token Audit skill is a powerful, open-source utility designed specifically for OpenClaw developers and operators who need to gain clear visibility into their LLM token usage. It scans workspace files and provides accurate token estimations across six major model families: GPT-4, Claude 3, Gemini, Llama 3, Mistral, and Qwen. By surfacing critical data regarding hidden cost drivers, this skill helps developers pinpoint bloated context, redundant files, and massive optimization opportunities.
Operating with zero external dependencies and requiring no API keys, this skill functions as a privacy-centric, standalone script. It ensures that you understand exactly how your data footprints map to token metrics and commercial hosting costs before sending payloads to LLM providers, effectively extending the lifetime and budget of your automated workflows.
Token Audit Use Cases
- Optimizing Agent Budgets: Identifying high-cost files within an agent workspace before launching large-scale automated tasks.
- Context Window Management: Preventing context overflow errors by auditing token counts across different LLM backends prior to execution.
- Workspace Clean-up: Discovering redundant or disproportionately large files that consume massive token counts relative to their developmental utility.
- CI/CD Quality Gates: Integrating token and cost estimation checks into local repository hooks to maintain lean codebases.
How Token Audit Works
- Workspace Scanning: The script targets either the default OpenClaw workspace directory or a user-defined custom path.
- Token Estimation: It processes every valid file in the workspace, evaluating token lengths tailored to tokenization characteristics across six major LLM providers.
- Cost Projection: The script applies current pricing models to the estimated token counts, generating clear financial projections per model.
- Report Generation: It surfaces recommendations, identifying files that are disproportionately large, and outputs the final metrics in human-readable terminal format or machine-readable JSON.
Token Audit Setup
To start using the Token Audit skill within your environment, run the script from your terminal. No external packages or API keys are required.
Scan the Default Workspace
python3 scripts/token-audit.py
Scan a Target Path
python3 scripts/token-audit.py --workspace /path/to/workspace
Export Data as JSON
python3 scripts/token-audit.py --json
Token Audit Data Schema & Taxonomy
The skill processes your local file layout and generates analytical metrics structured around the following categories:
| Data Point | Description |
|---|---|
| Total Token Count | Consolidated token count across all scanned files, customized per model family. |
| Per-File Breakdown | An ordered list of files, sorted descending by token size, to quickly locate content bloat. |
| Cost Estimation | Projected financial cost metrics computed for GPT-4, Claude 3, Gemini, Llama 3, Mistral, and Qwen. |
| Optimization Flags | Actionable recommendations pointing out files that have unusually high token costs vs expected value. |
Token Audit Advanced Features
- Multi-Model Parity: Provides concurrent token calculations across six distinct provider specifications in a single execution sweep.
- Structured JSON Engine: Emits fully structured JSON payloads using the
--jsonflag, making it seamless to integrate into automated developer dashboards, CI/CD pipelines, or custom scripting pipelines. - Caelguard Ecosystem Compatibility: Operates natively alongside other Openclaw Skills and security tools like shellguard-scanner to establish comprehensive local safety gates.
SKILL.md
Loading
Related Openclaw Skills
AI E-Commerce Mall and Conversational Checkout
An interactive conversational e-commerce skill that enables AI agents to guide users through product catalogs, manage shopping carts, calculate subtotals, and execute payments.
AI Tipping and Donation Skill
An AI-powered tipping and donation skill that intelligently recognizes user gifting intent, validates the transaction amount, and handles secure payment handoffs.
Alibaba Cloud DataWorks Open API
An AI agent skill designed to programmatically control and orchestrate Alibaba Cloud DataWorks resources via dynamic runtime API discovery and official SDKs.
SocratiCode MCP
An advanced MCP server that equips your AI agent with semantic code search, AST-aware codebase indexing, and live dependency graph analysis.
ClawHub Industry Trends (clawhub-top)
An automated trend analysis tool to monitor, rank, and report the most popular, trending, and newly released Openclaw Skills on ClawHub.
sher-deploy
A zero-config CLI tool that automatically detects, builds, and deploys frontend applications to instant, shareable live preview URLs.
Featured*
