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 MadAppGang/magus --skill madbench-evalsgit clone --depth 1 https://github.com/MadAppGang/magusWrote 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/madappgang/magus/madbench-evals)<a href="https://agentmods.dev/skills/madappgang/magus/madbench-evals"><img src="https://agentmods.dev/badge/skills/madappgang/magus/madbench-evals/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/madappgang/magus/madbench-evals"><img src="https://agentmods.dev/badge/skills/madappgang/magus/madbench-evals.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 | $0.00057 | $0.08158 |
| Opus 5.5 | $0.00023 | $0.03263 |
| Sonnet 5.5 | $0.00011 | $0.01632 |
| Haiku 4.5 | $0.00006 | $0.00816 |
Grade A, and why
madbench-evals 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 yesterday.
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 — 502 lines — stays where its author put it; the contents beside it link to each section on GitHub.
madbench Benches and Experiments — author, run, debug
madbench is a Go harness that benchmarks agentic coding tools (Claude Code, etc.). You write a
bench file (madbench.yaml / *.madbench.yaml), point it at a harness and model, give it
scenarios with seeded testdata, and grade the agent's work with checks.
These skills belong to madbench. They are written in the madbench repository and ship with the release they describe; any other copy is a copy. A mistake here is fixed upstream, never by editing a copy.
The madbench release these files were written for is in MADBENCH_VERSION beside this
file — one line, x.y.z. No file here restates a version number — five files each
asserting one are five places to fall out of step with the binary. Reference files:
glossary.md— read first, every time (Step 0): the terms, the deprecated and renamed terms, other tools' terms and what madbench calls eachschema.md— every bench and Experiment key: scenarios,repo:·setup:·generate:·image:·cwd:, the sandbox block,driver:,unproven:,metrics:expressions and the bench'smodule/, params and placeholders,guard_changes:, how an unknown key failschecks-catalog.md— every Check family, incl. theenvironment:*family (nine types), the 18session:*types (the Driver's four andsession:matchamong them), thread/outcome scoping, matchers,readout:, judges andvotesharness-and-sandbox.md— theharness:block's keys,interactive:and permission modes, sandbox levels, the environment probe and MCP preflight, the whole CLI, exit codesdebugging.md— error→cause map, the two controls, report-JSON analysis, tuning
Every madbench capability claim in these files carries a docs/<file>.md:<line> citation
into the madbench checkout's docs/. Never pkg/**/*.go: the checkout builds dev.
What ships with it
13 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.
- checks-catalog.md 67 KB
- debugging.md 30 KB
- examples/environment-family/madbench.yaml 2.3 KB
- examples/image-generated/generate.ts 971 B runs code
- examples/image-generated/madbench.yaml 1.5 KB
- examples/metrics-module/madbench.yaml 1.3 KB
- examples/metrics-module/module/index.ts 267 B runs code
- examples/metrics-module/module/package.json 73 B
- examples/plugin-staging/madbench.yaml 1.5 KB
- glossary.md 9.0 KB
- harness-and-sandbox.md 49 KB
- MADBENCH_VERSION 7 B
- schema.md 38 KB
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.
- yesterday Changed · +2 lines 2fac2f50dda0
- 3d ago Changed · +1 lines 458b11236867
- 10d ago Changed · +56 lines 214c50b0d4cd
- 20d ago Changed a3e94dab6251
- 1mo ago Changed 4768c11dabc5
- 1mo ago Changed · +170 lines · -3 tokens per session bacda143cc14
- 1mo ago First seen · 273 lines · 60 tokens per session scan A 48e1066a167d
madbench-evals is a skill published in the GitHub repository MadAppGang/magus (13 stars, last pushed today), licensed MIT. It adds 57 tokens to every session and 8,158 once invoked, about $0.0002 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 2026-09-04.
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.
eval-dataset-design
A guide to creating reliable test sets for AI agents. Each evaluation task includes the request, the simulated user's behaviour, a reset starting state, and a success check that can be independently verified.
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.