← Back to Skills Marketplace
449
Downloads
0
Stars
5
Active Installs
2
Versions
Install in OpenClaw
/install rtk-compress
Description
Save 60-90% of LLM tokens on shell commands, file reads, and test outputs. Wraps rtk CLI for compressed output.
Usage Guidance
Before installing or enabling this skill: 1) Verify the rtk project: review its repository, security/telemetry docs, and release artifacts (prefer a pinned release over 'master' install). 2) Be cautious with curl | sh installs; prefer package manager installs or verified release tarballs/checksums. 3) Understand what rtk's 'proxy' and 'tracking' features do — confirm whether command outputs or environment values are sent to remote endpoints and whether sensitive values are redacted. 4) Treat the skill as able to read files and env vars used by the agent; avoid using it on commands that may expose secrets unless you've audited rtk's redaction. 5) Ask the skill author (or check a trusted source) to update metadata to list rtk as a required binary and to document telemetry and exact install instructions (pinned release). If you cannot audit rtk, consider running it in an isolated environment or avoiding commands that surface secrets.
Capability Analysis
Type: OpenClaw Skill
Name: rtk-compress
Version: 1.15.2
The skill bundle promotes a CLI tool (rtk) using a high-risk installation method (curl | sh) and explicitly instructs the agent to use commands that target sensitive data, such as 'rtk env -f AWS' in SKILL.md. While these features are framed as token-saving optimizations, providing an agent with direct shortcuts to filter and display AWS credentials or other environment variables poses a significant risk of accidental or intentional credential exposure to the LLM context.
Capability Assessment
Purpose & Capability
The SKILL.md consistently describes a wrapper around the rtk CLI to compress shell and file outputs — that matches the advertised purpose. However the registry metadata declares no required binaries while the instructions assume the rtk binary is installed; the skill should have declared rtk as a required binary.
Instruction Scope
The runtime instructions tell the agent to run many rtk commands that read files, env vars (e.g., `rtk env -f AWS`), run network fetches (`rtk curl`, `rtk wget`), and use `rtk proxy <command>` which the doc labels as 'Raw passthrough + tracking'. Those behaviors are plausible for a token-compression wrapper, but they also mean command outputs, file contents, or filtered env values may be processed and potentially transmitted by rtk. The SKILL.md does not document telemetry or remote endpoints used by rtk, nor does it restrict what the agent should send externally.
Install Mechanism
This is an instruction-only skill with no install spec, but the doc instructs installing rtk via Homebrew or by piping a GitHub-hosted install.sh (curl | sh) from the repository's master branch. Installing via a shell pipe from an unpinned branch is higher-risk than a pinned release; although the URL is GitHub (a known host), invoking an unverified install script can execute arbitrary code on the host.
Credentials
The skill declares no required environment variables, which is reasonable for a CLI wrapper. But the instructions include commands that inspect environment variables (e.g., `rtk env -f AWS`) and read config files (e.g., `rtk json config.json`). That behavior is consistent with the stated purpose, but because env/config inspection can surface secrets, you should assume rtk may access and process sensitive values unless you confirm it strips or redacts them.
Persistence & Privilege
The skill does not request always:true, has no install manifest writing into system config, and makes no claims about modifying other skills. It appears to be an on-demand wrapper; autonomous invocation is allowed by default but is not a new privilege introduced by this skill.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install rtk-compress - After installation, invoke the skill by name or use
/rtk-compress - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.15.2
- Added a structured YAML header with metadata (name, version, description, tags).
- Expanded and detailed the Commands Reference section, including example usages for file, git, GitHub, test, build, lint, containers, analytics, and token savings features.
- Updated language throughout for clarity and completeness.
- Existing agent rules, savings reference table, and installation instructions retained.
v1.0.0
Initial release — save 60-90% of LLM tokens on exec commands using rtk
Metadata
Frequently Asked Questions
What is Rtk Compress?
Save 60-90% of LLM tokens on shell commands, file reads, and test outputs. Wraps rtk CLI for compressed output. It is an AI Agent Skill for Claude Code / OpenClaw, with 449 downloads so far.
How do I install Rtk Compress?
Run "/install rtk-compress" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Rtk Compress free?
Yes, Rtk Compress is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Rtk Compress support?
Rtk Compress is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Rtk Compress?
It is built and maintained by erscoder (@erscoder); the current version is v1.15.2.
More Skills