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.
curl -O https://raw.githubusercontent.com/genlayerlabs/subzeroclaw/main/.claude/skills/subzeroclaw-contribute/SKILL.mdgit clone --depth 1 https://github.com/genlayerlabs/subzeroclawWrote 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.
[](https://agentmods.dev/skills/genlayerlabs/subzeroclaw/subzeroclaw-contribute)<a href="https://agentmods.dev/skills/genlayerlabs/subzeroclaw/subzeroclaw-contribute"><img src="https://agentmods.dev/badge/skills/genlayerlabs/subzeroclaw/subzeroclaw-contribute/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/genlayerlabs/subzeroclaw/subzeroclaw-contribute"><img src="https://agentmods.dev/badge/skills/genlayerlabs/subzeroclaw/subzeroclaw-contribute.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00100 | $0.01087 |
| Opus 5 | $0.00050 | $0.00544 |
| Sonnet 5 | $0.00020 | $0.00217 |
| Haiku 4.5 | $0.00010 | $0.00109 |
Grade A, and why
subzeroclaw-contribute scanned grade A with 1 finding 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 10d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
`tee`, `curl` itself); plugin systems, hooks, event buses, middleware (multi-user How it starts
The opening of the file, as written. The whole thing — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Contributing to SubZeroClaw
SubZeroClaw is an anti-framework: the goal is not to grow but to stay minimal,
readable and correct. The whole runtime is one file, src/subzeroclaw.c
(~550 lines). Before you add anything, read CONTRIBUTING.md — a PR that adds
surface has to clear a high bar, and several categories are refused outright.
The two contributions that are welcome
- Reduce complexity — same behaviour with less code, fewer moving parts, or clearer logic: remove a function that doesn't earn its place, flatten a control path, drop a dependency or an allocation, replace a hand-rolled pattern with libc. A PR that removes lines while keeping tests green is probably good.
- Prove a runtime limitation — a task where "skill + shell + LLM loop" genuinely
breaks down in the runtime, not the skill. The honest test: could a better
.mdskill fix it? If yes, it's a skill problem, not a SubZeroClaw one. If the limitation is structural (e.g. binary tool output can't round-trip through the JSON string protocol), that's worth discussing. Open an issue with the task, the skill you wrote, what happened, and why no skill can work around it.
Won't be merged: new tools (the shell is the only tool — the model runs git,
tee, curl itself); plugin systems, hooks, event buses, middleware (multi-user
platform problems SubZeroClaw doesn't have); backward-compatibility shims.
Code map (src/subzeroclaw.c)
The loop, top to bottom:
| Function | Role |
|---|---|
config_parse_line / config_load |
parse the config keys; env overrides; scrub the four provider secrets (SUBZEROCLAW_API_KEY/_ENDPOINT/_REQUEST_EXTRA/_COMPACT_EXTRA) from the environment (wipe-in-place + unsetenv) so the model's shell never sees the key. SUBZEROCLAW_SKILLS is preserved on purpose. |
main |
read config + skills, build the system prompt, seed the message array, drive the loop. |
agent_run |
the turn loop: POST to endpoint, parse, dispatch tools, until a stop or max_turns. |
parse_response |
pull assistant text / tool calls / errors out of the completion JSON. |
process_tool_calls |
the one tool: run each shell command via popen() (stderr merged), append results as turns. |
round_has_command |
guard: did this round actually request a shell command? |
compact_fire / compact_splice / compact_url |
async context sealing: on the router's x_router.compact signal, fire /v1/compact in the background and splice the sealed block in ahead of intervening turns — no turn is ever blocked. |
log_write, write_temp, mkdirp |
session logging + scratch files. |
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.
- 10d ago First seen · 74 lines · 100 tokens per session scan A b5e3909052f3
subzeroclaw-contribute is a skill published in the GitHub repository genlayerlabs/subzeroclaw (134 stars, last pushed 4d ago), licensed MIT. It adds 100 tokens to every session and 1,087 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
goga-cell-cpp
C++ CODEMANIFEST contract implementation rules.
cpp_expert
Modern C++ specialist (C++17/20/23). Smart pointers, move semantics, and generic programming.
doca-argp
Use this skill for hands-on DOCA Arg Parser CLI work on a shipped sample or new DOCA-using app — adding / removing / renaming flags; wiring docaargpinit → register params → docaargpstart → docaargpdestroy in order; picking a parameter type from the full public enum (DOCAARGPTYPESTRING, INT, BOOLEAN, DEVICE, DEVICEREP…
cuopt-numerical-optimization-api
LP, MILP, and QP (beta) with cuOpt — Python, C, and CLI. Use when the user is solving LP, MILP, or QP with any cuOpt interface.
c-code-style
Write, edit, or review C (and C-compatible header) code according to C coding style rules, then check and run clang-format to enforce formatting. Trigger on "write C code", "edit a .c/.h file", "review C code style", "format C code", "run clang-format", "check clang-format", "apply the C style guide".
readable-cpp
Readable C/C++/Rust/CUDA code rules inspired by The Art of Readable Code. Use when writing, reviewing, or refactoring C, C++, Rust, or CUDA code. Enforces short functions, flat control flow, clear naming, readable structure, and idiomatic patterns.