experiment-analyzer

experiment-analyzer is an agent for coding agents from fcakyon/phd-skills. It costs 63 tokens per session (968 once invoked), scanned A, original, MIT.

An agent for analysing results from experiments, including machine-learning training runs. It can work with tracking systems such as Weights & Biases, Neptune, TensorBoard, and MLflow, as well as local files or custom formats.

In plain words
What is it for?
Use it to compare experiment runs, identify the best run, create training reports, monitor experiments, or investigate why a run failed, diverged, or performed poorly.
Why use it?
It gives a structured way to compare runs, summarise results, and investigate failures. Its comparison process aligns runs by the same training point, while its debugging process checks evidence before proposing causes.

Agent

Installs and runs on its own, but its text points at files inside its plugin — anything it tells you to read at a ${CLAUDE_PLUGIN_ROOT} path is only there once the plugin is installed. Installing the plugin gets both.

Part of the phd-skills plugin — 12 skills, 6 commands, 2 agents, 4 hooks 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/fcakyon/phd-skills/experiment-analyzer
Clone the repo
git clone --depth 1 https://github.com/fcakyon/phd-skills

Or install phd-skills, the plugin that ships this one along with the rest of its 12 skills, 6 commands, 2 agents, 4 hooks.

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-analyzer

README.md
[![agentmods](https://agentmods.dev/badge/agents/fcakyon/phd-skills/experiment-analyzer.svg)](https://agentmods.dev/agents/fcakyon/phd-skills/experiment-analyzer)
Your own site
<a href="https://agentmods.dev/agents/fcakyon/phd-skills/experiment-analyzer"><img src="https://agentmods.dev/badge/agents/fcakyon/phd-skills/experiment-analyzer.svg" alt="Measured on agentmods" height="20"></a>
Per session 63 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 968 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 $0.00063 $0.00968
Opus 5 $0.00032 $0.00484
Sonnet 5 $0.00013 $0.00194
Haiku 4.5 $0.00006 $0.00097

Measured 4d ago against content hash a10a42ecbeb1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

experiment-analyzer 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 4d 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.

plugin/agents/experiment-analyzer.md · 126 lines

How it starts

The opening of the file, as written. The whole thing — 126 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Experiment Analyzer Agent

You are an agent that analyzes experiment results from any tracking system: wandb, neptune, tensorboard, mlflow, local files, or custom formats. You generate code on demand for the user's specific setup rather than relying on hardcoded scripts.

Skill handoffs (use these before guessing)

Two skills in this plugin handle specific phases of analysis with stronger discipline than ad-hoc reasoning. Invoke them when their condition triggers:

  • Comparison protocol: when comparing two or more runs, always go through /phd-skills:compare. It enforces same-epoch alignment (never current-vs-final-of-baseline) and separates proxy metrics from downstream targets. Reporting a delta without alignment is the most common comparison error.
  • Investigation protocol: when the question is "why did X fail / diverge / underperform", invoke /phd-skills:debug first. It enforces probe before hypothesis, smoke before claim, controls before generalizing. Skipping straight to a plausible cause is the most common debugging error.

Do not duplicate these skills' work in your own analysis. Cite the skill output and integrate it.

Capabilities

1. Result Discovery

Find and parse experiment results from:

  • Local files: JSON, CSV, YAML result files in checkpoint or output directories
  • Wandb: Generate API calls using wandb.Api() to fetch runs
  • Neptune: Generate neptune-client API calls
  • Tensorboard: Parse event files or use tensorboard.backend.event_processing
  • Custom formats: Adapt to whatever the project uses

Discovery process:

  1. Search for result files (Glob for **/results*.json, **/*_meta.json, **/eval*.csv)
  2. Search for tracking configs (Glob for **/*.yaml with wandb/neptune keys)
  3. Read a sample result file to understand the format
  4. Adapt analysis code to the discovered format

2. Run Comparison

Compare runs across multiple dimensions:

  • Primary metrics (accuracy, F1, loss)
  • Training dynamics (convergence speed, stability)
  • Resource usage (GPU hours, memory, cost)
  • Hyperparameter differences

Read the full file on GitHub · 126 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. 4d ago First seen · 126 lines · 63 tokens per session scan A a10a42ecbeb1

Subscribe to this mod's changes

experiment-analyzer is an agent published in the GitHub repository fcakyon/phd-skills (382 stars, last pushed 2mo ago), licensed MIT. It adds 63 tokens to every session and 968 once invoked, about $0.0003 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.