feasibility-salvage-gate
用于 P4 判断实验计划可实施性、预计论文级别、low-confidence 风险、失败后贡献重定位和最低补救方案。不用于拒稿信预演。
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Menu
用于 P4 判断实验计划可实施性、预计论文级别、low-confidence 风险、失败后贡献重定位和最低补救方案。不用于拒稿信预演。
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
用于 P4 设计最低可交付 baseline、投稿级强 baseline、消融实验、指标阈值和 stop condition。不用于真正运行实验。
用于把论文贡献主张映射成实验矩阵、指标、成功判定、失败动作和机器 verifier 契约。不用于 baseline 细化或图表设计。
用于 P4 设计论文图、表、截图/系统演示证据,绑定 claim、章节位置、caption 和数据字段 schema。不用于实际画图。
Experiment-planning-suite 主入口。用于 P4 实验规划、实验矩阵、图表证据链、做一半能不能发、怎么交给 req 拆实验任务时路由。不用于真正跑实验(交给 req-suite 执行)。
用于把 knowledge/experiment-plan/ 的 P4 实验规划转成 req-suite 下一棒 prompt,要求拆 TASK-EXP-* 并绑定 verifier。不用于亲自创建任务文件。
用于 P4 实验规划前先读项目材料,生成研究主张快照、资源边界、预计论文级别和 low-confidence 缺口。不用于拆 req 任务。
| name | feasibility-salvage-gate |
| description | 用于 P4 判断实验计划可实施性、预计论文级别、low-confidence 风险、失败后贡献重定位和最低补救方案。不用于拒稿信预演。 |
本 skill 是 P4 的策略门。它判断当前计划能不能做、能支撑什么论文级别、哪里低置信,以及失败后怎样换论文策略。
/feasibility-salvage-gate(
snapshot,
experiment_matrix,
baseline_ablation_plan,
figure_table_plan,
target_level?,
output_root = knowledge/experiment-plan/
) -> feasibility_salvage_plan
按维度打标:
| Dimension | Question |
|---|---|
| data | 数据、案例、工况、截图是否可获取? |
| code/environment | 代码、工具、仿真、平台是否能跑? |
| baseline | 最低 baseline 和强 baseline 是否可实现? |
| metric | 指标是否可自动或半自动判定? |
| figure/table | 图表数据能否产出? |
| time/resource | 当前资源是否够做 must-have 实验? |
| paper_level | 证据链是否匹配目标级别? |
输出:
current_support_level: <band>
target_level: <band or unknown>
confidence: high|medium|low
blocking_gaps: [...]
upgrade_path: [...]
minimum_salvage_path: [...]
low-confidence 不是失败;它必须附带最低补救方案。
优先从论文策略上补救:
Write artifact expectations to:
knowledge/experiment-plan/machine/experiment-artifacts.schema.json
This schema should tell harness agents how to recognize results, plots, screenshots, metric files, and verifier outputs.
Write or update:
knowledge/experiment-plan/feasibility-salvage-plan.md
knowledge/experiment-plan/machine/experiment-artifacts.schema.json
Do not make rejection-letter simulation the core framing. The gate can mention likely reviewer concerns only when they directly map to evidence gaps, but the main frame is contribution repositioning and feasibility.
plugins/experiment-planning-suite/skills/feasibility-salvage-gate/,bump plugin 版本,刷新 marketplace/cache。