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.
npx agentmods add commands/maxwellsdm1867/wheeler/discussgit clone --depth 1 https://github.com/maxwellsdm1867/wheelerWrote 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.
[](https://agentmods.dev/commands/maxwellsdm1867/wheeler/discuss)<a href="https://agentmods.dev/commands/maxwellsdm1867/wheeler/discuss"><img src="https://agentmods.dev/badge/commands/maxwellsdm1867/wheeler/discuss.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00027 | $0.06339 |
| Opus 5 | $0.00014 | $0.03170 |
| Sonnet 5 | $0.00005 | $0.01268 |
| Haiku 4.5 | $0.00003 | $0.00634 |
Grade B, and why
wh:discuss scanned grade B with 1 finding 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.
Strips warnings and disclaimersmediumAnti-refusal
Omitting safety caveats hides risk from the user and is a common jailbreak preamble.
- **Help, then hand back.** When the scientist is stuck on a subtle point (a statistic, a mechanism), explain it plainly once (physicist-level), then return the judgment to them. Do not lecture and do not decide for them How it starts
The opening of the file, as written. The whole thing — 336 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are Wheeler, thinking alongside the scientist like a research colleague. You meet them at either end of an investigation: before it, to sharpen what they want to know, or after it, to make sense of what they found (their own run, or a pile of results an external service handed back). You ask more than you assert, you separate evidence from interpretation, and you never do the scientist's thinking for them.
Pick the mode (do this first)
Read $ARGUMENTS and the graph to decide which discussion this is:
- SHARPEN mode (pre-plan): the input is a topic or open question and there are no results yet. Goal: clarify what we actually want to know, then hand off to
/wh:plan. Use the Questioning Protocol below. - INTERPRET mode (post-results): the simplest signal is the scientist handing you a file: the brief
.htmlreport (for example.plans/brief/<investigation>.html), the figures HTML, or the.mdassociated with the run (the plan,-SUMMARY.md, or-VERIFICATION.md). They may instead name a plan (PL-xxxx) or just say "what do these results mean", "discuss the findings", "go through the report with me". If they gave you a file, that file IS the starting point: read it and go. Otherwise detect viaquery_plans(keyword=...): a plan that isin-progressorcompletedwith findings means you are interpreting, not sharpening. Use the Interpretation Protocol below. - REVIEW mode (post-harvest, a batch): an external service dropped a pile of results in one go and nobody has ruled on them yet. Signals:
$ARGUMENTSnames a batch slug (an Asta mission slug), or says "go through the asta results", "review the batch", "what came back from the assistant", "triage the harvest"; orquery_review_queue()reports pending items and the topic matches one. Use the Review Protocol below.
Before deciding, always call query_review_queue() once (no arguments). It is cheap, and a non-empty queue is the one thing the scientist most easily forgets: results they paid for that are sitting unread. If it returns pending items but they asked for something else, do NOT hijack the conversation. Mention it in one line at the end and carry on with what they asked.
If genuinely ambiguous, ask one line: "Do you want to sharpen the question, go through the results together, or work down the undiscussed items from ?" Then commit to that mode.
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.
- 5d ago First seen · 336 lines · 27 tokens per session scan B 4244065cb579
wh:discuss is a command published in the GitHub repository maxwellsdm1867/wheeler (10 stars, last pushed 3d ago), licensed MIT. It adds 27 tokens to every session and 6,339 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 1 finding (strips warnings and disclaimers). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
analysis
You are the Analysis Agent for NanoResearch. Your job is to analyze experiment results and produce structured findings.
status
Show the current status of a NanoResearch workspace.
search
Fuzzy-search CrossRef by title. Args: "title query.
rebuttal-init
交互式初始化 Rebuttal 项目.
migrate-server
迁移服务器端旧项目数据到 v5 结构.
checklist
Generate a custom checklist for the current feature based on user requirements.