菜单
Run one autoresearch cycle — hypothesize, Director review, execute on Modal, evaluate. Use to manually trigger a research cycle or to test the loop.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Close out an experiment — update run card with results, update frontmatter, rebuild docs index, close PR, log to research log, propose program.md updates. Use when an experiment has finished training and eval.
Pre-flight checklist for launching a training experiment on Modal — verifies run card, config, data contract, local smoke test, budget, observability, and stop conditions before burning GPU time. Use when kicking off any new experiment, Mamba2 or JEPA. Counterpart to /experiment-complete.
Get Mamba2 and training architecture guidance for experiment design. Use when a proposed experiment touches model structure, training regime, or loss design and you need to understand how the architecture constrains or enables the approach.
Search the world model research literature for prior art, alternative approaches, or techniques from adjacent fields. Use when evaluating whether a proposed technique has precedent, when seeking inspiration from different paradigms, or when a paper is referenced and you need context on it.
Review project experiment history to answer questions about what's been tried, what worked, hit rates, and empirical patterns. Use when proposing a new experiment direction and you need to check whether it's been explored, or when evaluating results against prior findings.
Run one conductor heartbeat — check experiment state, evaluate completed runs, dispatch new research cycles. Supports parallel experiments. Use with `/loop 60m /conductor` for autonomous operation.
基于 SOC 职业分类
| name | research-cycle |
| description | Run one autoresearch cycle — hypothesize, Director review, execute on Modal, evaluate. Use to manually trigger a research cycle or to test the loop. |
| disable-model-invocation | true |
| user-invocable | true |
| argument-hint | [scout|confirm|dry-run] |
Run one cycle of the autoresearch loop. Can be run manually or by the simple-loop conductor.
/research-cycle or /research-cycle scout — full cycle, Scout tier (<$2)/research-cycle confirm — full cycle, Confirm tier ($2-10)/research-cycle dry-run — hypothesis + Director review only, no Modal executionRead .loop/state/budget.json. If daily or weekly limit would be exceeded, report and stop.
Spawn an Explore agent with the hypothesis agent prompt (.loop/agents/hypothesis.md). It reads program.md and recent cards, returns a structured hypothesis + draft run card.
Spawn an Explore agent with the research-director prompt (.loop/agents/research-director.md). Pass it the hypothesis. It returns APPROVE / REVISE / REJECT with reasoning.
Present the hypothesis AND the Director's review to the user. If dry-run mode, stop here.
If REJECT: log to RESEARCH-LOG.md. Done.
Ask the user for confirmation before spending money: "Director approved [experiment]. Estimated cost: $X. Launch on Modal?"
If confirmed:
modal run --detach/experiment-complete when the run finishes..loop/state/log.jsonl.loop/state/budget.json