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 wenhaochai/claude-plugins --skill convert-to-benchmark-scriptgit clone --depth 1 https://github.com/wenhaochai/claude-pluginsWrote 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/wenhaochai/claude-plugins/convert-to-benchmark-script)<a href="https://agentmods.dev/skills/wenhaochai/claude-plugins/convert-to-benchmark-script"><img src="https://agentmods.dev/badge/skills/wenhaochai/claude-plugins/convert-to-benchmark-script/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/wenhaochai/claude-plugins/convert-to-benchmark-script"><img src="https://agentmods.dev/badge/skills/wenhaochai/claude-plugins/convert-to-benchmark-script.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 | $0.00099 | $0.01222 |
| Opus 5.5 | $0.00040 | $0.00489 |
| Sonnet 5.5 | $0.00020 | $0.00244 |
| Haiku 4.5 | $0.00010 | $0.00122 |
Grade A, and why
convert-to-benchmark-script 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 14d 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 — 92 lines — stays where its author put it; the contents beside it link to each section on GitHub.
convert-to-benchmark-script
In: the script the user actually runs.
Out: a new script that does the same job within a few minutes (mostly by reducing total steps), gives the same answer twice, and leaves nothing behind.
Four rules
- It has to be the same job. Same model size, batch size, sequence length, precision, parallelism, and world size etc. Shrink any of them and the bottleneck moves somewhere else, and then the loop spends a day optimizing the wrong thing.
- It has to be short. Aim for one run within 8 minutes. Every candidate in the loop pays this cost, so a 20 minute benchmark is a 20 minute tax on every experiment. Short is a budget, not an instruction to cut: if the job already fits, leave it whole.
- It has to leave nothing behind. No runs on someone's dashboard, no checkpoints eating a quota, no writes into the checkout.
- It has to give the same answer twice. Pin every seed and every source of
ordering. Use the script's own flag where it has one, and otherwise seed from
your wrapper before the workload starts.
optimize's setup smoke-runs it 3 times and reads the tolerance off those runs, so drift you leave in here comes back there as a tolerance too wide to gate anything on.
What to turn off
Reach for the script's own flag or an environment variable every time. A deleted line is a change to their code that nobody agreed to.
Estimate the cost before you switch anything off. You cannot measure it here, so do the arithmetic on paper:
- a checkpoint:
bytes on disk / write bandwidth x saves per run - an eval pass:
eval batches x 1/3 of a train step / steps between evals
Divide either by the run's total wall clock. Under a few percent, turn it off. Above that, either measure it as a phase of its own rather than buried in steady-state step time, or turn it off and record the exclusion, with your estimate, in the contract's Anything else row. An exclusion nobody wrote down becomes a speedup that never shows up in the real job.
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.
- 14d ago Changed · +35 tokens per session e3cc8c22bcc3
- 17d ago First seen · 92 lines · 64 tokens per session scan A 3788dcaa1c25
convert-to-benchmark-script is a skill published in the GitHub repository wenhaochai/claude-plugins (17 stars, last pushed 2d ago), licensed MIT. It adds 99 tokens to every session and 1,222 once invoked, about $0.0004 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 2026-10-01.
Other skills, from other repositories
cli-eval
Create and run evaluation suites, watch live benchmark progress, view scorecards, compare model performance, and integrate eval runs with CI workflows from the CLI.
model-merging
Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task…
dummy-dataset
Generate realistic dummy datasets for testing with customizable columns, constraints, and output formats (CSV, JSON, SQL, Python script). Use when creating test data, building mock datasets, or generating sample data for development and demos.
darwinian-evolver
Evolve prompts/regex/SQL/code with Imbue's evolution loop.
bad-case-to-dpo
A guide to turning failed production agent runs into training and evaluation data. It explains how to find the first mistake, create preference examples for DPO (Direct Preference Optimization), and check that a fix does not damage other tasks.
validate
Validate Semantica pipelines, extraction quality, graph schemas, and ontology consistency. Returns structured error/warning checklists. Uses PipelineValidator, PipelineBuilder.validatepipeline(), GraphValidator, and OntologyValidator. Sub-commands: pipeline, step, dependencies, extraction, graph, ontology, performance.