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 madaeroblade/atlas --skill atlas-benchmarkgit clone --depth 1 https://github.com/madaeroblade/atlasWrote 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/madaeroblade/atlas/atlas-benchmark)<a href="https://agentmods.dev/skills/madaeroblade/atlas/atlas-benchmark"><img src="https://agentmods.dev/badge/skills/madaeroblade/atlas/atlas-benchmark/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/madaeroblade/atlas/atlas-benchmark"><img src="https://agentmods.dev/badge/skills/madaeroblade/atlas/atlas-benchmark.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 | $0.01026 |
| Opus 5 | $0.00017 | $0.00513 |
| Sonnet 5 | $0.00007 | $0.00205 |
| Haiku 4.5 | $0.00003 | $0.00103 |
Grade A, and why
atlas-benchmark 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 9d 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 — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Atlas Benchmark
Resolve Atlas
Read and follow:
$ATLAS_ROOT/engine/benchmark/ENGINE.md$ATLAS_ROOT/engine/benchmark/SCORING.md
Use repository fixtures from:
.atlas/benchmark/scenarios/.atlas/benchmark/metrics.md.atlas/benchmark/coverage.yaml
If benchmark fixtures do not exist, report that Atlas benchmarking is not initialized.
Critical rule
You are the orchestrator. You never answer a scenario yourself.
Reading .atlas/INDEX.md contaminates this context permanently. Any scenario
you answer here is an Atlas-on result with no control, however carefully you
try to set the knowledge aside.
Benchmarking is also read-only: do not repair .atlas/ before or during
scoring. Measure first, repair later.
Execution
Every scenario runs as three subagents.
Use a subagent type that starts from a fresh context. A forked or context-inheriting agent carries your contamination into the arm and silently invalidates the run.
1. Atlas-off executor
Give it: the scenario prompt verbatim, and the repository.
Instruct it: .atlas/ is out of scope — do not read it, do not list it, do not
grep it. Work from source, tests, config and Git history.
Also instruct it that the Atlas section of the project CLAUDE.md does not
apply to it and must be disregarded. A subagent inherits the project
instruction file, and in a bootstrapped repository that file tells it to read
.atlas/INDEX.md on every non-trivial task. Without this, the control arm
arrives pre-contaminated and the run measures nothing.
Do not tell it a knowledge layer exists beyond that exclusion.
Then check the returned transcript. If it read, listed or grepped .atlas/
anyway, the scenario is VOID — report it, and do not score it.
2. Atlas-on executor
Give it: the same prompt verbatim, and the repository.
Instruct it: this repository has an Atlas at .atlas/; use it as its own
INDEX.md describes.
3. Scorer
Give it: the scenario file, and both transcripts labeled A and B, without saying which arm produced which.
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.
- 9d ago First seen · 134 lines · 35 tokens per session scan A 6410d47d12e4
atlas-benchmark is a skill published in the GitHub repository madaeroblade/atlas (2 stars, last pushed 29d ago), licensed MIT. It adds 35 tokens to every session and 1,026 once invoked, about $0.0002 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 2026-08-31.
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".
neuron-evaluation-engineer
Create and run AI evaluations with datasets, assertions, and output drivers in Neuron AI. Use this skill whenever the user mentions evaluation, testing AI systems, creating evaluators, dataset-driven testing, assertion-based validation, or wants to measure AI system performance. Also trigger for tasks involving…
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.