Borrowing it
Nothing to install: this file belongs to openclaw/shellbench. 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/openclaw/shellbench/main/.agents/skills/shellbench-research-runbook/SKILL.mdgit clone --depth 1 https://github.com/openclaw/shellbenchWrote 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/openclaw/shellbench/shellbench-research-runbook)<a href="https://agentmods.dev/skills/openclaw/shellbench/shellbench-research-runbook"><img src="https://agentmods.dev/badge/skills/openclaw/shellbench/shellbench-research-runbook/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/openclaw/shellbench/shellbench-research-runbook"><img src="https://agentmods.dev/badge/skills/openclaw/shellbench/shellbench-research-runbook.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 | $0.00076 | $0.01124 |
| Opus 5.5 | $0.00030 | $0.00450 |
| Sonnet 5.5 | $0.00015 | $0.00225 |
| Haiku 4.5 | $0.00008 | $0.00112 |
Grade A, and why
shellbench-research-runbook 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 2mo 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 — 111 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ShellBench Research Runbook
Use this skill for a real benchmark campaign, not a one-off local score.
Read references/runbook.md before provisioning machines. It is the normative campaign contract and contains the commands, gates, artifact schema, and recovery rules.
Non-negotiable gates
- Use remote Crabbox AWS beasts for benchmark execution. Never run scored trials on the operator laptop.
- Pin one public-task commit, runner commit or patch hash, provider model ID, harness version, reasoning level, and judge route for the whole campaign.
- Run one
r0qualification for every distinct harness and model-family route, using exactly ten pinned representative tasks. Do not start full-suite jobs until model identity, real traces, tools, usage, judge routing, and artifact export pass. - Retain and audit every
r0, but force it out of leaderboard scoring. Qualify with independent full-suite repetitionsr1throughr3. After a clean audit, addr4throughr6; the research result is six total repetitions. - Run every provider-supported non-maximum reasoning level. Never label a reasoning level as tested unless the route applies it and the trace or proxy evidence proves it. Record unsupported levels instead of fabricating them.
- Use
gpt-5.6-solathighas the default judge. Keep the judge alias, credentials, logs, and identity audit separate from the agent route. - Start checkpointing after the first completed trial and continue at least every ten minutes or ten new results. Verify each local archive before it counts.
- Upload every verified checkpoint and final archive to the private S3 prefix
from
SHELLBENCH_TRACE_S3_URI. Never put bucket names or credentials in git, PR text, public logs, or generated reports. - A run is not research-clean when traces are missing, observed model identity differs from the request, reasoning is unproven, coverage is incomplete, or infrastructure failures dominate.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 2mo ago First seen · 111 lines · 76 tokens per session scan A 7c327fcaaae0
shellbench-research-runbook is a skill published in the GitHub repository openclaw/shellbench (141 stars, last pushed 23d ago), licensed MIT. It adds 76 tokens to every session and 1,124 once invoked, about $0.0003 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 2026-08-30.
Other skills, from other repositories
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.
tika-eval-compare
Compare extracts from two Tika builds over a corpus to detect regressions in content, encoding, exceptions, and embedded-document handling. Use for "compare before/after extracts", "eval this change against the corpus".
cua-driver
Use Cua Driver for desktop or browser tasks that are awkward or unavailable through Bash/APIs, or when the user explicitly wants GUI interaction: app testing, visual bug reproduction, form filling, calendar entry, screenshots, and demo recording. Also covers Cua setup; not OpenAI Codex Computer Use or web research.
jetson-validate-image
Use after jetson-flash-image to run static BSP checks, on-target smoke/regression tests on a flashed DUT, or both. Not for build or flash steps. Triggers: validate bsp, on-target validation.
atmos-validation
Validate Atmos projects, components, arbitrary JSON Schema inputs, EditorConfig, and GitHub Actions; use affected-file selection and native CI annotations.
skill-benchmark
Benchmark AI skill effectiveness by measuring implementation quality against legacy constraints.