experiment-interpreter

experiment-interpreter is an agent for coding agents from ai-analyst-lab/ai-analyst-plugin. It costs 38 tokens per session (1,535 once invoked), scanned A, original, MIT.

An agent that classifies experiment results as Ship, Abort, Learn, or Invalid using pre-registered decision rules. It checks whether the experiment and its data are trustworthy before interpreting the outcome.

In plain words
What is it for?
Use it to interpret analysis output together with an experiment.yaml file and decide whether to release the tested change, stop it, learn from it, or reject the results.
Why use it?
It reduces after-the-fact explanations by applying the planned rules consistently, including checks for broken randomization, missing data, and implementation bugs.

Agent

Part of the ai-analyst-plus plugin — 44 skills, 1 command, 13 agents shipped together

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/ai-analyst-lab/ai-analyst-plugin/experiment-interpreter
Clone the repo
git clone --depth 1 https://github.com/ai-analyst-lab/ai-analyst-plugin

Or install ai-analyst-plus, the plugin that ships this one along with the rest of its 44 skills, 1 command, 13 agents.

Wrote 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.

agentmods badge for experiment-interpreter

README.md
[![agentmods](https://agentmods.dev/badge/agents/ai-analyst-lab/ai-analyst-plugin/experiment-interpreter.svg)](https://agentmods.dev/agents/ai-analyst-lab/ai-analyst-plugin/experiment-interpreter)
Your own site
<a href="https://agentmods.dev/agents/ai-analyst-lab/ai-analyst-plugin/experiment-interpreter"><img src="https://agentmods.dev/badge/agents/ai-analyst-lab/ai-analyst-plugin/experiment-interpreter.svg" alt="Measured on agentmods" height="20"></a>
Per session 38 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,535 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 $0.00038 $0.01535
Opus 5 $0.00019 $0.00767
Sonnet 5 $0.00008 $0.00307
Haiku 4.5 $0.00004 $0.00153

Measured 5d ago against content hash 63f9370dec5c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

experiment-interpreter 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 5d 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.

Origin

Copies of this mod

1 near-identical copy found in the catalogue:

ai-analyst-plus/agents/experiment-interpreter.md · 155 lines

How it starts

The opening of the file, as written. The whole thing — 155 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent: Experiment Interpreter

Purpose

Classify experiment outcomes using a structured decision framework. Takes raw analysis results and pre-registered decision rules, walks the Result Interpretation Tree, and produces one of four verdicts: Ship, Abort, Learn, or Invalid. Prevents post-hoc rationalization by anchoring every decision to pre-registered criteria.

Inputs

  • {{ANALYSIS_RESULTS}}: Path to analysis output (JSON or markdown from Experiment Analyzer).
  • {{EXPERIMENT_CONFIG}}: Path to experiment.yaml with pre-registered decision rules.

Framework: Result Interpretation Tree

Branch 1: Check Validity First

Before interpreting results, verify the experiment itself was valid:

SRM verdict?
├── BLOCK → INVALID: Randomization broken. Cannot trust any results.
├── WARNING → Flag but continue with caution.
└── PASS → Proceed to interpretation.

Data quality issues?
├── >10% missing outcomes → INVALID: Too much missing data.
├── Implementation bug detected → INVALID: Treatment didn't deploy correctly.
└── Clean → Proceed.

If INVALID: Stop. Do not interpret. Report what went wrong and recommend fixes.

Branch 2: Interpret Primary Metric

Primary metric result?
├── Significant POSITIVE (p < alpha, lift > 0):
│   └── Check guardrails → Branch 3
├── Significant NEGATIVE (p < alpha, lift < 0):
│   └── ABORT: Treatment hurt the primary metric.
├── Not significant (p >= alpha):
│   ├── Was the experiment adequately powered (≥80%)?
│   │   ├── YES → ABORT: Powered null. No evidence of benefit.
│   │   │   Note: "The experiment had sufficient power to detect a
│   │   │   [MDE] effect. Observing no significant effect means the
│   │   │   true effect is likely smaller than [MDE]."
│   │   └── NO → LEARN: Underpowered null. Effect may exist but
│   │       we couldn't detect it.
│   │       Recommendations:
│   │       - Extend the experiment
│   │       - Increase traffic allocation
│   │       - Choose a more sensitive metric
│   │       - Accept the inconclusive result and move on
│   └── Compute the CI. If CI includes practically meaningful effects,
│       flag: "We cannot rule out a [X]% effect."

Read the full file on GitHub · 155 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. 5d ago First seen · 155 lines · 38 tokens per session scan A 63f9370dec5c

Subscribe to this mod's changes

experiment-interpreter is an agent published in the GitHub repository ai-analyst-lab/ai-analyst-plugin (32 stars, last pushed 9d ago), licensed MIT. It adds 38 tokens to every session and 1,535 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 2026-08-30.

Related

Other agents, from other repositories

editor

Journal editor who desk-reviews manuscripts, selects two referees with deliberately different dispositions, calibrates to a target journal from .claude/references/journal-profiles.md, and synthesizes an editorial decision (FATAL / ADDRESSABLE / TASTE). Used by /review-paper --peer [journal].

pedrohcgs/claude-code-my-workflow · 64 tokens

Geoprocessing Specialist

ArcPy and Python toolbox expert who automates spatial workflows — builds .pyt toolboxes, Model Builder processes, batch geoprocessing automation, and custom analysis scripts for ArcGIS Pro.

SHAdd0WTAka/Zen-Ai-Pentest · 45 tokens

research-scout

Scans the NeqSim codebase to discover scientific paper opportunities that will drive code improvement. Every paper must improve NeqSim — adding tests, validating models against data, hardening algorithms, or implementing new capabilities. Produces ranked, actionable topics that feed into the planner agent.

equinor/neqsim · 61 tokens

algorithm-expert

RL algorithm expert. Fire when working on GRPO/PPO/DAPO/GSPO/SAPO algorithms, reward functions, advantage normalization, loss computation, or training loop implementation.

redai-infra/Relax · 37 tokens

mathodology-problem-analyst

Use for contest problem decomposition, scoring criteria, constraints, variables, assumptions, and deliverable mapping.

sweetcornna/mathodology · 29 tokens

astronomical-instrumentation-scientist

Reasons from system-level error budgets, the diffraction limit and Strehl ratio, detector figures of merit, and resolving power through Zemax/Code V tolerancing, ETC radiometry, AO modeling, and on-sky standard-star commissioning while treating flexure drift, IR persistence, ghosts, and quasi-static speckles as…

K-Dense-AI/scientific-agents · 78 tokens