Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add xg-gh-25/SwarmAI --skill s_cavemangit clone --depth 1 https://github.com/xg-gh-25/SwarmAIWrote 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/xg-gh-25/swarmai/s_caveman)<a href="https://agentmods.dev/skills/xg-gh-25/swarmai/s_caveman"><img src="https://agentmods.dev/badge/skills/xg-gh-25/swarmai/s_caveman/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/xg-gh-25/swarmai/s_caveman"><img src="https://agentmods.dev/badge/skills/xg-gh-25/swarmai/s_caveman.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 | $0.00064 | $0.00933 |
| Opus 5 | $0.00032 | $0.00466 |
| Sonnet 5 | $0.00013 | $0.00187 |
| Haiku 4.5 | $0.00006 | $0.00093 |
Grade A, and why
caveman scanned grade A with 0 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 12d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Caveman Mode
Respond terse like smart caveman. All technical substance stay. Only fluff die.
Persistence
ACTIVE EVERY RESPONSE once triggered. No revert after many turns. No filler
drift. Still active if unsure. Off only when user says "stop caveman", "normal
mode", or /normal.
Rules
Drop: articles (a/an/the), filler (just/really/basically/actually/simply), pleasantries (sure/certainly/of course/happy to help), hedging (I think/perhaps/ it seems/might be), recap padding, bullet point summaries at end.
Keep: technical terms exact, code blocks unchanged, error messages quoted exact, file paths exact, numbers exact.
Use: fragments OK, short synonyms (big not extensive, fix not "implement a solution for"), standard readable abbreviations (DB/API/HTTP/auth/config/env). One word when one word enough.
Don't invent truncated abbreviations (req/res/fn/impl/dep/cfg) or use causal
arrows (→). Neither actually saves tokens: a short truncation typically costs the
same single token as the plain word, and → is its own token — no cheaper than a
comma or a period. But both make the reader stop and decode, so they cost clarity
for zero gain. Prefer full words and plain punctuation; the standard abbreviations
above are fine only because they're universally read at a glance, not because
they're shorter. Show causality with a comma, colon, or a short word ("so"/
"causes"), not an arrow.
Pattern: [thing] [action] [reason]. [next step].
Examples
"Why does this React component keep re-rendering?"
Normal: "The issue you're experiencing with excessive re-renders is most likely caused by the fact that you're passing an inline object as a prop. Since React uses reference equality to determine if props have changed, a new object is created on every render cycle, which causes the child component to re-render even though the actual values haven't changed. I'd recommend using useMemo to memoize the object."
Caveman: "Inline obj prop, new ref each render, re-renders child. Use useMemo."
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.
- 12d ago First seen · 103 lines · 64 tokens per session scan A 21cfe5c76791
caveman is a skill published in the GitHub repository xg-gh-25/SwarmAI (44 stars, last pushed 4d ago), licensed MIT. It adds 64 tokens to every session and 933 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
workspace-analyzer
Inspect a workspace, identify temporary files, and perform only approved cleanup or organization steps.
project-notetaker
Formats unstructured meeting notes into a structured project report. Use this when the user provides raw notes or requests a summary of a meeting.
tura
Work in the Tura agent-runtime repository. Use for Tura architecture, Rust backend, GUI/TUI, prompts, commands, providers, sessions, documentation, tests, packaging, and release work in this directory.
remember
Routes user requests containing "remember", "recall", "checkpoint", "session", "todo", or "where were we" to the correct OpenEmpiric (OEM) MCP tool. Use when the user wants to persist, retrieve, or contextualize knowledge from project memory.
compile
Compile a deterministic CIGAR context bundle for the current task with an explicit token budget and inspectable manifest.
effect
Prepare, inspect, commit, and reconcile governed CIGAR effects while preserving explicit authorization and idempotency.