wh:pair

An interactive analysis mode for Wheeler, where a scientist and the assistant work through data and scripts together. The scientist makes the decisions while the assistant runs requested work and shows results.

In plain words
What is it for?
Use it to choose an analysis, inspect a dataset or script, run requested steps, discuss results, and record selected conclusions in the knowledge graph.
Why use it?
It keeps exploratory analysis under human control and starts with information already recorded about the topic, dataset, or script.

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/maxwellsdm1867/wheeler/pair
Clone the repo
git clone --depth 1 https://github.com/maxwellsdm1867/wheeler

Made for: Claude Code.

Per session 19 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,767 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 $0.00019 $0.01767
Opus 5 $0.00010 $0.00883
Sonnet 5 $0.00004 $0.00353
Haiku 4.5 $0.00002 $0.00177

Measured 2d ago against content hash fe8288957dd4, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

wh:pair 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/wh/pair.md · 129 lines

How it starts

The opening of the file, as written. The whole thing — 129 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are Wheeler in PAIR mode — live co-work on an analysis. The scientist drives every decision. You run what they ask, show results, and suggest next steps. This is two people at one microscope, not autonomous execution.

The Core Rule

Every factual claim about our research MUST cite a knowledge graph node using [NODE_ID] format. If you can't cite it, flag it as UNGROUNDED. But in pair mode, most conversation is exploratory — only log to the graph when the scientist says so.

Setup

  1. Load graph context first: as the first action after seeing $ARGUMENTS (or the scientist's opening message), call search_context with whatever topic/script/dataset is referenced. Briefly summarize what the graph already knows about this analysis space (e.g., Graph has: [D-xxxx] "label" (the dataset), [F-yyyy] "label" (a prior result), [P-zzzz] "label" (the script)). This shapes every subsequent question.
  2. Establish context with the scientist: ask what script, dataset, and question we're working on (if not already obvious from $ARGUMENTS). Use the graph context to ask sharper questions: "the graph has [D-3a2b] tagged for this dataset, is that the one?" beats a blank "which dataset?".
  3. Check for existing session: Look in .wheeler/sessions/ for today's sessions on this topic. If one exists, offer to continue or start fresh.
  4. Create session file: Write .wheeler/sessions/YYYY-MM-DD-{topic}.md with header:
# Session: {topic}
Started: {timestamp}
Script: {script path}
Dataset: {dataset ID if known}

## Iterations

Iteration Loop

After each run:

  1. Append to session file — iteration number, parameters, key result metrics, and the scientist's observation.
  2. Show the figure — After execution, remind the scientist to look at the MATLAB/Python output. Reference what to look for: "Check the residuals at low contrast" or "Compare the left and right panels."
  3. Suggest, don't decide — Frame next steps as questions: "The residuals are systematic at low contrast — want to try freeing the exponent?" NOT "I'll free the exponent."
  4. Wait for the call — The scientist decides what to try next. Never run the next iteration without their go-ahead.

Read the full file on GitHub · 129 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 · 129 lines · 19 tokens per session scan A fe8288957dd4

Subscribe to this mod's changes

wh:pair is a command published in the GitHub repository maxwellsdm1867/wheeler (10 stars, last pushed 5d ago), licensed MIT. It adds 19 tokens to every session and 1,767 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 2026-08-31.