briefbound-ai-research-loop skillA
briefbound-ai-research-loop is agent-read markdown (skill) from ccdawn/codex-skills: Use when AI/ML research engineering needs baseline reproduction, hypothesis-driven experiments, ablations, evaluation design, findings synthesis, research direction selection, plateau recovery, or a reproducible loop from a paper/repository/research question to evidence-backed results, or coordinating cross-stage research competitions and benchmarks..
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
# AI 研究工程循环 ## 目标 把 AI 研究从“不断尝试代码”变成两个相互连接的轻量循环： ```text 内层：可信 baseline -> 可证伪假设 -> 最小实验 -> 评估 -> 接受/拒绝/保留 外层：汇总多轮证据 -> 提炼规律 -> 更新假设组合 -> 继续/分支/转向/停止 ``` 本 skill 是 AI 研究工程的主 owner。`briefbound-score-loop` 只承接一条可量化实验 lane；跨阶段竞赛/benchmark 按需读 [竞赛阶段参考](references/competition-lifecycle.md)。 ## Briefbound task contract - Context Boundary: 研究问题、代码和数据来源、active baseline、评价协议、允许修改面、计算预算、实验记录和当前证据。 - Output Contract: baseline 复现结论、假设组合、实验 lane、证据综合、研究方向决策或可复现交接。 - Allowed Action: 在已锁定的可编辑面和预算内复现、修改、运行、评估和记录；不静默改变数据划分、metric、baseline 或研究目标。 - Success Evidence: 可复现命令、baseline 指纹、metric 与方差、diff/config、实验 artifact、对照/消融结果以及有来源的研究结论。 - Stop Condition: baseline 不可信、评价协议漂移、数据泄漏、预算或权限不足、结果不可复现、关键假设无法区分，或继续实验已无新的信息价值。 - Route Out: `briefbound-score-loop`、`briefbound-feature-reuse-research`、`briefbound-bug-review`、`briefbound-research-rigor-review`、完成交接或 BLOCKED。 - Support 图解：研究汇报复杂时自动加载 `briefbound-diagram-design` 配图（不夺所有权）；短状态不配图。 ## 统一调用契约 - 只处理 Briefbound task contract 范围；不匹配时回 `briefbound-router` 或更具体 owner，复合任务不吞其他 owner。 …
Read the whole file at its exact version.
How to install
mdr add ccdawn/codex-skills/briefbound-ai-research-loop@git:20260917.27f1ad2mdr add ccdawn/codex-skills/briefbound-ai-research-loop@sha256:70e83a41ab8fa14cPin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_ybzbh77wohfy4kvp)
1 badge views in 30 days
Versions
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20260917.27f1ad2 latest | 2026-09-17 | 27f1ad2 | 7,957 B | A | view · diff |
| git:20260917.0939a38 | 2026-09-17 | 0939a38 | 7,984 B | A | view · diff |
| git:20260916.292b0ac | 2026-09-16 | 292b0ac | 7,824 B | A | view · diff |
| git:20260813.dec1d14 | 2026-08-13 | dec1d14 | 7,610 B | A | view |
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (7957 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
ccdawn/codex-skills · 6 stars · license MIT · pushed 2026-09-22 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_ybzbh77wohfy4kvp GET https://markdownregistry.com/api/v1/resolve?ref=ccdawn/codex-skills/briefbound-ai-research-loop GET https://markdownregistry.com/api/v1/blob/70e83a41ab8fa14c05692dff8445a6ce392b10aeed1583e3979da08c1c691b78
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.