subzeroclaw: Skill for Claude Code

.claude/skills/subzeroclaw-use/SKILL.md

subzeroclaw-use is a skill for Claude Code from genlayerlabs/subzeroclaw. It costs 109 tokens per session (1,513 once invoked), scanned D, original, MIT.

Instructions for running SubZeroClaw, a small C program that repeatedly asks a language model for work and runs its shell commands. It can load markdown files as agent instructions and use installed command-line tools.

In plain words
What is it for?
Use it to configure SubZeroClaw, add an agent instruction file, run tasks, connect it to a router, or run it as a system service.
Why use it?
It explains how to build, configure, and operate the program while warning that model output runs directly on the machine without confirmation or a sandbox.

Skill for Claude Code

Written for Claude Code: installed under .claude/. Also seen: mentions Claude Code.

This is genlayerlabs/subzeroclaw's own configuration. It tells Claude Code how to work on subzeroclaw itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything subzeroclaw configures →

Reuse

Borrowing it

Nothing to install: this file belongs to genlayerlabs/subzeroclaw. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/genlayerlabs/subzeroclaw/main/.claude/skills/subzeroclaw-use/SKILL.md
Clone the repo
git clone --depth 1 https://github.com/genlayerlabs/subzeroclaw

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for subzeroclaw-use

README.md
[![agentmods](https://agentmods.dev/badge/skills/genlayerlabs/subzeroclaw/subzeroclaw-use.svg)](https://agentmods.dev/skills/genlayerlabs/subzeroclaw/subzeroclaw-use)
Your own site
<a href="https://agentmods.dev/skills/genlayerlabs/subzeroclaw/subzeroclaw-use"><img src="https://agentmods.dev/badge/skills/genlayerlabs/subzeroclaw/subzeroclaw-use.svg" alt="Measured on agentmods" height="20"></a>
Per session 109 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,513 The whole file, excluding the scripts and references it only reads on demand.
Security scan D 3 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00109 $0.01513
Opus 5 $0.00055 $0.00757
Sonnet 5 $0.00022 $0.00303
Haiku 4.5 $0.00011 $0.00151

Measured 8d ago against content hash 4f1c78d8b27a, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade D, and why

subzeroclaw-use scanned grade D with 3 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Asks for rootmediumPrivilege escalation

A mod that escalates privileges can change anything on the machine, not only the project.

`Restart=on-failure` and an `EnvironmentFile` (root-owned, `chmod 600`) holding

Recursive force deletehighDestructive command

rm -rf with a variable or a broad path is one typo away from removing the wrong tree.

`popen()` — no confirmation, no sandbox, `rm -rf /` included. There is nothing

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

loop until done**. One tool: `shell`. The "adapter" for git/curl/email/ffmpeg is
.claude/skills/subzeroclaw-use/SKILL.md · 111 lines

How it starts

The opening of the file, as written. The whole thing — 111 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Using SubZeroClaw

Read the warning first. SubZeroClaw runs whatever the model emits through popen() — no confirmation, no sandbox, rm -rf / included. There is nothing between the model's output and your system. Run it only where you accept that, ideally as an unprivileged service user (see As a service).

The whole runtime is: read a skill (markdown) → call an LLM → run shell tools → loop until done. One tool: shell. The "adapter" for git/curl/email/ffmpeg is that they're one popen() away — install the CLI, the model uses it.

Two senses of "skill" — don't conflate them. A SubZeroClaw skill is a plain .md file in ~/.subzeroclaw/skills/ that becomes the agent's system prompt (what the agent knows). This file is a Claude Code skill (how to operate the tool). When the docs below say "write a skill", they mean the former.

Build & run

git clone https://github.com/genlayerlabs/subzeroclaw && cd subzeroclaw
make                                   # ~0.5s → the 55KB binary (needs gcc; curl at runtime)
mkdir -p ~/.subzeroclaw/skills
# ...write ~/.subzeroclaw/config (below) and a skill .md...
./subzeroclaw "check disk usage and clean tmp if over 80%"   # one-shot
./subzeroclaw                                                 # interactive
make install                           # → ~/.local/bin/

An agent skill is just prose the LLM reads — no format, no registry, no trigger matching. Drop a .md in ~/.subzeroclaw/skills/ and it joins the system prompt. The repo's skills/*.md are format examples tied to one setup; write your own.

Configuration (~/.subzeroclaw/config)

These are the only keys the runtime parses (config_parse_line); config.example is the canonical template.

Key Meaning
api_key The unhardcoded router consumer key (llmr_…) — SubZeroClaw is designed to run on the router. A bare OpenRouter/provider key also works but is a degraded loop. Required.
request_extra JSON merged into every request body. Carries the model ({"model":"..."}) — there is no dedicated model key. Against a router it also carries the routing policy_ir. Override wins on key collision.
compact_extra JSON enabling async compaction: keep_recent + a cheap summariser policy_ir. Unset → no compaction.
endpoint default OpenRouter; point at an unhardcoded router for routing/cache/compaction.
skills_dir, log_dir defaults under ~/.subzeroclaw/.
max_turns tool-call loops per input (default 200).

Read the full file on GitHub · 111 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 8d ago First seen · 111 lines · 109 tokens per session scan D 4f1c78d8b27a

Subscribe to this mod's changes

subzeroclaw-use is a skill published in the GitHub repository genlayerlabs/subzeroclaw (134 stars, last pushed 2d ago), licensed MIT. It adds 109 tokens to every session and 1,513 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it D with 3 findings (asks for root, recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

research-repository

Build a repository that makes findings findable, reusable, and cumulative across teams. Use when the same research keeps getting redone. For synthesising one study, use affinity-diagram.

Owl-Listener/designer-skills · 43 tokens

design-negotiation

Advocate for design quality, scope, and timeline with partners and leadership using evidence and shared goals. Use in the conversation itself. For the commercial vocabulary behind it, use business-design (ux-strategy).

Owl-Listener/designer-skills · 48 tokens

conversational-ux

Design voice and conversational interfaces — dialog flows, error recovery, and persona. Use when the interface speaks and listens rather than being tapped. For graphical input collection, use form-design.

Owl-Listener/designer-skills · 43 tokens

platform-conventions

Design to iOS and Android conventions — what each OS mandates, where they diverge, and when to unify. Use when shipping native apps. For breakpoint adaptation use responsive-design; for matching competitor patterns use jakobs-law (interaction-design).

Owl-Listener/designer-skills · 55 tokens

law-of-similarity

Apply the Law of Similarity — shared colour, shape, or size signals that elements belong to one category. Use when signalling relationships across distance. For grouping by position, use law-of-proximity.

Owl-Listener/designer-skills · 47 tokens

form-design

Design a form end to end — field order, grouping, validation, and completion. Use when the artifact is a form. For product-wide error strategy use error-handling-ux; for first-run signup use onboarding-design.

Owl-Listener/designer-skills · 51 tokens