一键导入
genomeft-run-campaign
Use when supervising an official GenomeHarness v15 campaign or sequential suite through the interactive CLI 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。
菜单
Use when supervising an official GenomeHarness v15 campaign or sequential suite through the interactive CLI 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。
Use when the user naturally asks to start, continue, monitor, or summarize a GenomeHarness campaign or sequential benchmark suite.
Use when a GenomeHarness v15 campaign has pending PROPOSE requests and proposal JSON must be written from request packets.
Use when a GenomeHarness campaign has a REPAIR request and repair JSON must be written without changing scientific protocol.
Use when generating or reviewing GenomeHarness v15 campaign reports from state.sqlite and trial artifacts.
基于 SOC 职业分类
| name | genomeft-run-campaign |
| description | Use when supervising an official GenomeHarness v15 campaign or sequential suite through the interactive CLI loop. |
Use plans/genomeharness_v15_plan.md as the active authority.
genharness plan --dry-run preview.genharness preflight --campaign campaigns/X/config.yaml.genharness start --campaign campaigns/X/config.yaml.genharness advance --campaign campaigns/X --wait.PROPOSE, read pending packet(s) and use the proposal skill.FREEZE, read the FREEZE packet and submit a freeze JSON selecting a completed node.REPAIR, use the repair skill.DONE, run genharness report --campaign campaigns/X; the report command deletes checkpoint directories by default after the report is written.FAILED, report the failure and do not claim final gain.advance --wait is only a client attach. The durable runner owns subprocesses and GPU leases.
For a sequential NT/GB suite:
genharness suite-init --suite suites/X --model ... --benchmark-family nt.genharness suite-start-next --suite suites/X.DONE or FAILED.genharness suite-start-next --suite suites/X again for the next task.The suite starts one task at a time. Do not run multiple benchmark tasks in parallel unless the user explicitly changes that design.
advance --wait attach.genharness status --campaign ... --brief, then genharness advance --campaign ... --wait.