gpu-analytic-decision-prior
Use when DeepSci decision work needs GPU analytical-modeling priors for routing mismatches, missing evidence, or unsupported claims.
- License 许可
- MIT
- 最近更新
- 2026-09-07
- Stars
- ★ 2
- Forks
- 0
01 前言 Front Matter
- name
- gpu-analytic-decision-prior
- description
- Use when DeepSci decision work needs GPU analytical-modeling priors for routing mismatches, missing evidence, or unsupported claims.
02 资源文件 2
- 参考文档agents/openai.yaml362 B
- 参考文档commands/mismatch-route.md700 B
03 SKILL.md 正文
GPU Analytic Decision Prior
Overview
This prior helps route GPU analytical-modeling work when model predictions, measurements, profiler evidence, or claims do not line up.
Workflow
- When explicitly requested during decision work, use
mismatch-routeto classify the smallest unresolved mismatch. - Choose the smallest honest route: refine, add evidence, limit, park, block, or accept partial support.
- Record what evidence or repair would change the decision.
Subcommands
| Subcommand | Use For | Detail |
|---|---|---|
mismatch-route |
Decision prior for failed or incomplete model evidence | commands/mismatch-route.md |
Chat Response
Present normal chat responses in natural-language Markdown. Lead with the outcome, use descriptive headings when they improve readability, and use lists only for genuinely distinct items. Treat named output items as information to cover, not as literal response keys. Do not emit snake_case: value, pseudo-JSON, pseudo-YAML, or a flat program-style record unless the user explicitly requests machine-readable output. Keep exact schemas in durable artifacts and summarize them naturally in chat.
04 同仓库更多 Skills
fixture-shared-template
暂无简介
deepsci-org-shared-template
Generated shared guardrails for all deepsci-org Domain Agent Team Template participants; use with generated deepsci-org event and tick skills.
deepsci-org-operator-control
Generated operator-control skill for deepsci-org lifecycle semantics, mode switches, manual stepping, pause, resume, recovery, and stop intent.
deepsci-org-on-tick
Generated bounded tick skill for deepsci-org scheduling, recovery, completion, and manual-step decisions.
deepsci-org-on-team-start
Generated on-event skill for mail with schema_id deepsci-org.email.team-start; wakes deepsci-org-master for one bounded start or resume pass.
deepsci-org-on-handoff-result
Generated on-event skill for master mail with schema_id deepsci-org.email.handoff-result.