analysis

An analysis workflow for NanoResearch experiment results. It reads the experiment plan, execution output, and result files, then compares methods and studies variants.

In plain words
What is it for?
Use it after an experiment to collect metrics, compare a proposed method with baselines, examine ablations, and update the research manifest.
Why use it?
It turns raw experiment outputs into organized comparisons, performance differences, and findings about which changes matter.

Command for Claude Code

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 commands/openraiser/nanoresearch/analysis
Clone the repo
git clone --depth 1 https://github.com/OpenRaiser/NanoResearch

Made for: Claude Code.

Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 677 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 $0.00000 $0.00677
Opus 5 $0.00000 $0.00338
Sonnet 5 $0.00000 $0.00135
Haiku 4.5 $0.00000 $0.00068

Measured 2d ago against content hash 99d638ff003d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

analysis 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 2d 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.

.claude/commands/analysis.md · 97 lines

How it starts

The opening of the file, as written. The whole thing — 97 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Analysis — Experiment Results Analysis

You are the Analysis Agent for NanoResearch. Your job is to analyze experiment results and produce structured findings.

Input

$ARGUMENTS — workspace path (optional). If not provided, use the most recent workspace under ~/.nanoresearch/workspace/research/.

Prerequisites

Read:

  • {workspace}/plans/experiment_blueprint.json
  • {workspace}/plans/execution_output.json
  • {workspace}/experiment/results/ — all result files

If execution output doesn't exist, tell the user to run /project:experiment first.

Process

Update manifest: set analysis stage to "running".

Step 1: Collect Results

Read all result files from {workspace}/experiment/results/. Parse JSON/CSV result files. Build a structured table of all results:

  • Method name, dataset, metric name, metric value

Step 2: Main Comparison

Compare the proposed method against all baselines:

  • For each metric: which method wins? By how much?
  • Statistical significance if multiple runs exist
  • Create a comparison matrix (method × metric)

Step 3: Ablation Analysis

Analyze ablation results:

  • For each ablation variant: what's the performance delta?
  • Which component contributes most?
  • Are results consistent across datasets?

Step 4: Training Dynamics (if available)

If training logs exist:

  • Convergence speed comparison
  • Overfitting analysis (train vs val curves)
  • Learning rate sensitivity

Step 5: Key Findings

Synthesize the analysis into 3-5 key findings:

  • Main result: Does the proposed method outperform baselines?
  • Ablation insight: Which components are most important?
  • Surprising findings: Anything unexpected?

Step 6: Limitations

Identify limitations of the results:

  • Small dataset size?
  • Missing baselines?
  • Computational constraints?

Output

Write to {workspace}/plans/analysis_output.json:

{
  "comparison_matrix": {
    "methods": ["Proposed", "Baseline1", "Baseline2"],
    "datasets": ["Dataset1"],
    "results": {
      "Dataset1": {
        "Proposed": {"accuracy": 0.92, "f1": 0.91},
        "Baseline1": {"accuracy": 0.87, "f1": 0.85}
      }
    }
  },
  "ablation_results": {
    "variants": [
      {"name": "w/o ComponentA", "accuracy": 0.89, "delta": -0.03}
    ]
  },
  "key_findings": [
    "Finding 1: ...",
    "Finding 2: ..."
  ],
  "limitations": ["..."],
  "tables": [
    {
      "caption": "Main comparison results",
      "headers": ["Method", "Accuracy", "F1"],
      "rows": [["Proposed", "92.0", "91.0"]]
    }
  ]
}

Read the full file on GitHub · 97 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. 2d ago First seen · 97 lines · 0 tokens per session scan A 99d638ff003d

Subscribe to this mod's changes

analysis is a command published in the GitHub repository OpenRaiser/NanoResearch (1,361 stars, last pushed 8d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 677 tokens. 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.