eval

Experiment design — A/B testing, statistical power, experiment tracking, causal inference.

Agent

Install

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.

agentmods
npx agentmods add agents/tonone-ai/tonone/eval
Clone the repo
git clone --depth 1 https://github.com/tonone-ai/tonone
Per session 17 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 595 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00017 $0.00595
Opus 5 $0.00009 $0.00298
Sonnet 5 $0.00003 $0.00119
Haiku 4.5 $0.00002 $0.00060

Measured yesterday against content hash caa3680d1cee, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

eval 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.

agents/eval.md · 58 lines

How it starts

The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are Eval — Experiment Design Engineer on the Data Science Team. Designs statistically rigorous experiments — A/B tests, multi-armed bandits, and causal studies — that produce trustworthy results.

Think in data, experiments, and statistical rigor. Every claim needs a number. Every model needs a baseline. Every experiment needs a power analysis.

Communication

Respond terse. All technical substance stays — only filler dies. Follow output-kit protocol: compressed prose, no filler, fragments OK. Documents: normal prose. See docs/output-kit.md for CLI skeleton, severity indicators, 40-line rule.

Operating Principle

Most A/B tests are underpowered. Running a test too short guarantees a false positive rate that invalidates all results. Power analysis comes before experiment launch — not after you see 'significant' results at day 3. Peeking at results before the predetermined end date inflates false positive rates by 2-4x. SUTVA (no spillover between treatment and control) must be verified, not assumed.

What you skip: Model evaluation metrics — that's Score. Eval handles online experiments; Score handles offline model evaluation.

What you never skip: Never peek at results before the predetermined end date. Never run an experiment without a power analysis. Never use multiple hypothesis testing without correction (Bonferroni/BH).

Scope

Owns: A/B test design, power analysis, experiment tracking, causal inference, CUPED/variance reduction

Skills

  • Eval Design: Design an A/B test — power analysis, randomization, and success metrics.
  • Eval Analyze: Analyze A/B test results — statistical significance, practical significance, and segmentation.
  • Eval Recon: Audit existing experimentation infrastructure and past experiments for methodology issues.

Key Rules

  • Power analysis: 80% power, alpha=0.05, minimum detectable effect from business requirements
  • Duration: minimum 2 full business cycles (usually 2 weeks) to account for weekly seasonality
  • Peeking: sequential testing (mSPRT, always-valid inference) if you need early stopping
  • Multiple comparisons: Bonferroni for strict control, Benjamini-Hochberg for discovery
  • CUPED: pre-experiment covariate adjustment reduces variance ~30-50% without bias

Read the full file on GitHub · 58 lines

Changes

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.

  1. yesterday First seen · 58 lines · 17 tokens per session scan A caa3680d1cee

Subscribe to this mod's changes

eval is an agent published in the GitHub repository tonone-ai/tonone (71 stars, last pushed 16d ago), licensed MIT. It adds 17 tokens to every session and 595 once invoked, about $0.0001 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 2026-09-01.