ClawHub

token-kill

v0.0.1

Reduce OpenClaw token consumption by 95%+ using three optimization techniques (slash commands, script-first principle, and model tiering)

⭐ 0· 90·0 current·0 all-time
byTimmy@niuben
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
The name and description promise token-cost reductions and the SKILL.md contains only recommendations (slash commands, script-first, model tiering) that directly align with that goal. There are no unrelated required binaries, env vars, or installs.
✓
Instruction Scope
All runtime instructions are procedural recommendations (use /new, /compress, prefer scripts, route workloads to cheaper models). The file references (e.g., memory.md) are conceptual and the skill does not instruct the agent to read arbitrary files, exfiltrate data, or call unexpected external endpoints.
✓
Install Mechanism
No install spec or code files are included; this is an instruction-only skill, which minimizes on-disk/remote-code risk.
✓
Credentials
The skill requests no environment variables or credentials. Its recommendations to use external scripts imply that scripts will hold any needed secrets, which is proportionate to the guidance provided.
✓
Persistence & Privilege
The skill does not request always:true or other elevated persistence. It is user-invocable and uses the platform's normal invocation model; no extra privileges are requested.
Assessment
This skill is guidance-only and appears coherent and low-risk. Before adopting: verify that the platform slash commands (/compress, /new, /stop, /restart) actually behave as described and won't delete or irreversibly prune needed data; when you implement the recommended scripts, store any service credentials securely (not in model prompts); test model-tier routing in a sandbox to ensure quality/latency trade-offs are acceptable; and run a billing/audit check after changes to confirm expected savings. If you need stronger guarantees, ask the skill author for concrete examples or scripts that you can review.

Like a lobster shell, security has layers — review code before you run it.

latestvk977ts2zx1988m1k7b04gdzeqn83645c

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments