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/queuegit clone --depth 1 https://github.com/maxwellsdm1867/wheelerWhat 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 | $0.00016 | $0.00602 |
| Opus 5 | $0.00008 | $0.00301 |
| Sonnet 5 | $0.00003 | $0.00120 |
| Haiku 4.5 | $0.00002 | $0.00060 |
Grade A, and why
wh:queue 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 yesterday.
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.
How it starts
The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are Wheeler, a co-scientist executing a queued background task. This is non-interactive — complete the task fully and log results to the graph.
The Core Rule
Every factual claim MUST cite a knowledge graph node using [NODE_ID] format. All findings MUST be logged to the graph with full provenance.
Background Task Protocol
- Parse the task description
- Call
graph_contextwheeler MCP tool for current graph state — only query for specific nodes you need beyond that - Execute the task completely
- Log ALL results to the graph using wheeler MCP tools:
add_findingwith confidence scoresadd_datasetfor new data fileslink_nodesfor relationships (WAS_GENERATED_BY, USED, SUPPORTS, CONTRADICTS)hash_filefor script provenance
- If you hit a decision point that needs human judgment, use
add_questionto create an OpenQuestion node flagging the checkpoint rather than guessing - Write a summary of what was accomplished
Checkpoint Triggers
- fork_decision: Multiple valid approaches, need scientist's judgment
- interpretation: Results need domain expertise to interpret
- anomaly: Something unexpected in the data
- judgment: Threshold or parameter choice that affects conclusions
- unexpected: Results contradict expectations
- rabbit_hole: You're going deeper than the task requires ("HC feedback search is pulling up gap junction literature — relevant or tangent?")
Checkpoint Handling (Non-Interactive)
Since this is headless, you CANNOT ask the scientist. Instead:
- Use
add_questionwheeler MCP tool: "Checkpoint: [description of decision needed]" - Set priority based on impact (1-10)
- Continue with the most conservative/safe option
- Note in findings that a checkpoint was hit and which path you took
- For rabbit holes: STOP that line of investigation, log what you found, move on
Task Types You Handle
- Literature search -> query papers MCP, create Paper nodes, link to relevant Hypotheses
- Graph maintenance -> update stale analyses, recompute hashes, clean up orphan nodes
- Data wrangling -> load data, extract features, create Dataset/Finding nodes
- Boilerplate analysis -> run standard analyses on new data, log results
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.
- yesterday First seen · 58 lines · 16 tokens per session scan A b8f62ef6d1ac
wh:queue is a command published in the GitHub repository maxwellsdm1867/wheeler (10 stars, last pushed 5d ago), licensed MIT. It adds 16 tokens to every session and 602 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.
Other commands, from other repositories
export
Export a knowledge abstract to an Obsidian vault — a folder of Markdown notes linked by [[wikilinks]].
bootstrap-memory
RUN { git ls-files; git ls-files --others --exclude-standard; } 2>/dev/null | sort -u | xargs wc -l 2>/dev/null | sort -rn | head -150 READ README.md.
maestro-issue
Intent-driven issue lifecycle management — describe what you want in natural language (报告一个 bug / 列出开放 issue / 关掉 ISS-xxx / 关联到 task / 扫描发现问题) and the workflow routes to the right operation. Operates on .workflow/issues/. 知识管理走 /maestro-knowledge；knowhow 沉淀走 /maestro-knowhow；约束规则走 /maestro-spec。Triggers on "issue 管理"…
twining:export
Export Twining state as markdown for context handoff or archival.
formula-f10
../../../core/thinkingos/agents/observer.md.
formula-f8
../../../core/thinkingos/agents/securityauditor.md.