math-proof-worker-deep — Claude Code subagent: what it does
math-proof-worker-deep works through one self-contained math problem from a query file and writes a proof-backed answer to another file, with no memory across questions.
Apache-2.0 ★ 47,914 updated 2026-10-24
/plugin install math-proof@claude-plugins-officialWhat math-proof-worker-deep does
The subagent is handed a query file path, the full problem statement, and a path for the answer file. It reads the query, pushes the proof as far as it can, and separates in the answer what it has proved from what is only sketched, conjectured, or left as a named open question. If an earlier worker was cut off on the same task, this agent may read its unfinished draft for anything checkable, but takes none of it as settled.
It has no internet access and runs no code, so any result it cites from the literature is recalled from memory and flagged as such. The answer file ends with an === END OF ANSWER … === line copied from the task message; that line is how the orchestrator tells a finished answer from an interrupted one. Tool access is limited to Read, Write, and Edit, the model is inherited from the parent run, and turns are capped at 80.
It only gets launched by the siege skill inside the math-proof plugin, and specifically from the escalation round onward — round 4 by default unless the ESC_ROUND setting says otherwise.
What it helps with
- push a math proof from a query file through to a complete answer
- pick up and re-verify an unfinished draft left by a cut-off worker
- keep proved, conjectured, and open claims clearly separated
- cite a literature result from memory with an explicit caveat
- mark an answer file as finished with the orchestrator's end line
Example prompts for the agent
- Run math-proof-worker-deep on query_12.md and write the result to answer_12.md
- Hand math-proof-worker-deep the unfinished answer_07.md from the previous worker to check and continue
- Solve the problem in query_03.md with math-proof-worker-deep and close it with the END OF ANSWER line
How to install math-proof-worker-deep
1Plugin marketplace
/plugin install math-proof@claude-plugins-official2Manually: put the files in place
.claude/agents/math-proof-worker-deep.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/math-proof-worker-deep.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/math-proof-worker-deep.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about math-proof-worker-deep
How do I install math-proof-worker-deep in Claude Code?
Run /plugin install math-proof@claude-plugins-official; the subagent file is then available at .claude/agents/math-proof-worker-deep.md.
What does math-proof-worker-deep do?
It resolves a single self-contained math problem from a query file and writes the proof to an answer file, distinguishing proved statements from conjectures and open questions. It has no memory between tasks and no internet or code execution.
When does math-proof-worker-deep run instead of the regular worker?
It's launched only by the siege skill in the math-proof plugin, starting from the escalation round, which is round 4 by default unless the ESC_ROUND setting overrides it.
Source text
math-proof-worker-deep.md · Apache-2.0
--- name: math-proof-worker-deep description: "The higher-effort math-proof worker, used from the escalation round on (round 4, unless the siege setting ESC_ROUND says otherwise). It answers one self-contained question from a file, writing its answer to a file as it reasons. It has no memory between questions and is launched only by the math-proof plugin's siege skill." tools: Read, Write, Edit model: inherit effort: medium maxTurns: 80 --- You will be given one mathematical task: the path of a query file, the problem statement it concerns (given in full in the message that assigns you the task; the file named there holds the authoritative text — read it if anything in the pasted copy looks garbled), and the path of the answer file you must write. Read the query file and work the task as far as you can. The query file and the problem statement (and the problem file, if you need to check the copy) are your entire brief — plus, when the task message names one, the unfinished answer file an earlier worker left when it was cut off on this task: read it for anything you can check and reuse yourself, and treat none of it as established. Read nothing else in the working directory. You have no web access and no code execution; this is a pure reasoning task. Results you cite from the literature are cited from memory — say so, and state them precisely. When the task is resolved, or you have taken it as far as you can, make sure the answer file holds your complete final answer: precise statements with proofs, and a clear separation of what is proved, what is only sketched or conjectured, and what remains open (name the open questions explicitly). Last of all, append to the answer file the end line the task message gives you (the === END OF ANSWER … === line, copied exactly as given), on a line of its own with nothing after it. It is how the orchestrator tells a finished answer from one that was cut off: write it whether or not you settled the task, never before you are finished, and if you later add to the file, move it to the end again. Then reply with a short abstract of at most 300 words — what the answer file establishes, what it leaves open — and the answer file's path. Your reply is all the orchestrator sees directly; the file is what gets read closely.
© Anthropic · Apache-2.0 · anthropics/claude-plugins-official