Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
v2.4.0: Added shell aliases for auto-filtering (gst, dps, kpods, ok wrapper)
What This Skill Does
oktk sits between CLI command execution and the LLM context window, compressing verbose outputs from tools like git, docker, and kubectl into compact summaries. It intercepts command output, applies command-specific filters, and returns only the information the model needs. An AI learning mode handles commands that lack built-in filter rules.
Manually truncating CLI output before sending it to an LLM requires extra effort; oktk applies command-specific compression rules automatically, with a fallback to raw output if a filter fails.
When to use it
- Cutting API costs when repeatedly checking git status during a coding session
- Reviewing Docker container logs with an AI without flooding its context window
- Getting kubectl pod status summaries during cluster troubleshooting
- Running npm tests through an AI assistant without sending the full output
- Filtering curl responses to status codes and key headers during API debugging
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: What's the current state of my git repo?
- 1Runs git status in the project directory
- 2Pipes output through the oktk git status filter
- 3oktk extracts branch name, ahead/behind count, and file change counts
- 4Returns an 80-token compressed summary instead of the 800-token raw output
📍 main ↑ Ahead 3 commits ✏️ Modified: 6 ❓ Untracked: 3