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 roarista/awesome-harness --skill cavemangit clone --depth 1 https://github.com/roarista/awesome-harnessWrote 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/roarista/awesome-harness/caveman)<a href="https://agentmods.dev/skills/roarista/awesome-harness/caveman"><img src="https://agentmods.dev/badge/skills/roarista/awesome-harness/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/roarista/awesome-harness/caveman"><img src="https://agentmods.dev/badge/skills/roarista/awesome-harness/caveman.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00035 | $0.00355 |
| Opus 5.5 | $0.00014 | $0.00142 |
| Sonnet 5.5 | $0.00007 | $0.00071 |
| Haiku 4.5 | $0.00003 | $0.00036 |
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 1mo 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.
What it actually says
Caveman discipline
Keep the main session small: orchestrate, inspect evidence, and report the outcome. This is a behavioral workflow, not a claim of mechanical role enforcement.
For a bounded multi-file or non-trivial change:
- Write an atomic contract:
CONTEXT,CHANGE,GOAL,VERIFY. - Delegate that one contract to a builder. Do not recursively fan out.
- Give a distinct, read-only auditor the same contract. It independently runs the verification and returns PASS or FAIL with evidence.
- Keep the actual verification command and its exit code (R2) in the handoff/final evidence.
Do not delegate a genuinely trivial one-line edit; make the smallest safe change and leave a runnable check when the logic is non-trivial.
When graphify-out/graph.json exists, use graphify query or graphify explain before codebase exploration; after code changes, run graphify update .. Run the repository's factory/check-all gate before and after a change whenever it provides one.
At intermediate milestones, emit only caveman when the surface requires progress messages; otherwise emit nothing. The final response is self-contained.
Boundary
Native hooks cannot safely distinguish a main orchestrator from a delegated builder, nor cover every hosted tool, prose, or delegation action. Direct patches, model chat, and delegation therefore cannot be safely role-gated mechanically. Treat the workflow and its recorded verification as the enforcement boundary; use Git hooks as a commit-time backstop.
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.
- 1mo ago First seen · 26 lines · 35 tokens per session scan A 99f288829442
caveman is a skill published in the GitHub repository roarista/awesome-harness (1 stars, last pushed 4d ago), licensed MIT. It adds 35 tokens to every session and 355 once invoked, about $0.0001 per session on Opus 5.5. 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 2026-09-16.
caveman compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| cavemanroarista/awesome-harness | Skill | 1 | 4d ago | MIT | A | 35 | Claude Code, Codex |
| autoreviewopenclaw/clawhub | Skill | 9.5k | today | MIT | A | 17 | Claude Code, Codex |
| agent-teams-simplify-and-hardenLeoYeAI/openclaw-master-skills | Skill | 2.2k | 6d ago | MIT | A | 115 | Claude Code, Codex |
| omh-code-reviewrlaope/oh-my-hermes | Skill | 5.2k | today | MIT | A | 48 | Claude Code, Codex |
| revdiff-planumputun/revdiff | Skill | 947 | 2d ago | MIT | A | 84 | Claude Code |
| explore-unknownsdzhng/skills | Skill | 1.4k | today | MIT | A | 162 | Claude Code, Codex |
| re0-mergeLilMGenius/paperthin | Skill | 1.4k | 4d ago | MIT | A | 70 | Claude Code |