← All creators

a-green-hand-jack

User

Agent-native ML research repo template for AI PhD workflows with Claude Code: repo-as-control-plane, repo-local subagents/skills/hooks, anti-drift ANATOMY + runnable governance validators.

22 indexed · 0 Featured · 0 stars · avg score 53
Prolific

Categories

Indexed Skills (22)

AI & Automation Listed

adopt-existing-repo

把一个已有 Git repo 分步、可验证、尽量无人值守地收敛成本模板的完整形态;适用于迁移旧 ML/research repo,而不是从 GitHub template 新建空项目。

0 Updated 1 weeks ago
a-green-hand-jack
Data & Documents Listed

artifact-indexing

当有新的 dataset/checkpoint/table/figure 等 asset 产生或需要盘点时,用来维护 artifact 索引并出 stale asset report;绝不删或移动 bytes。

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

bootstrap-project

把一个刚从本模板派生(Use this template / gh repo create --template / clone+reinit)的新 repo 落地成自洽状态——写 .template.toml 版本锚点、启用 core.hooksPath、同步 Codex adapters、跑 governance;不猜测需要 human 信息的步骤,只报告待办。

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

coding-agent-quota

Read local Codex and Claude Code quota/rate-limit snapshots and produce routing evidence for agent work allocation, plus outcome-aware routing on top of quota (historical success/rework/latency evidence, deterministic offline replay, degraded fallback, outcome ledger with decision_id traceability). Use when deciding which coding agent/provider/model to use, balancing Codex vs Claude Code load, checking remaining 5-hour/session and weekly quota, estimating routing capacity before launching subagents, comparing candidate routes on quota + outcome evidence, recording route decisions/outcomes, or preparing cross-provider experiments with frozen provider/model policy.

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

experiment-workflow

当要跑一个实验或复现一篇 paper 时,用来写 experiment card、登记状态机 ledger、准备(但不擅自运行)launch 命令、只读监控与 alert、校验并交接 human 恢复提案、写 run summary 并把达标结果收进 evidence。

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

spawn

当要唤起一个 subagent(in-session 跑完汇报,或起一个持久可交互的 Paseo 标签页 agent)时用;串起「发现有哪些 agent → 选型/packet → 人+主 agent 两层交代 → 出生即好名并入 roster」,替代凭直觉直接开 general-purpose worker。

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

subagent-routing

当要为一个 child task 派 subagent 时,用来选定 model/effort/tools 并产出 launch packet;不要凭直觉直接开 general-purpose worker。

0 Updated 1 weeks ago
a-green-hand-jack
Data & Documents Listed

template-stress-test

用一个真实、非平凡的 case(外部项目或真实场景)压力测试模板自身的 validators/hooks/skills/subagent/command,在模板大改之后或想验证某类机制是否真的可用时使用;测试记录与修复实现分离,绝不在 case 分支里顺手改模板。

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

worktree-pr-flow

当要落地一处代码改动时,用来走 issue→branch→worktree→实现→更新anatomy/ledger→测试+validator→PR→review→merge→归档的标准流程;push/PR/merge 走 human gate。

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

anatomy-drift-control

在改动 repo 结构(新增/改名/移动文件)前后,用来 grep 所有引用并在同一 commit 更新 anatomy,防止结构地图漂移。

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

command-adopt-existing-repo

Codex adapter for Claude slash command /adopt-existing-repo - 分步迁移已有 Git repo,使其收敛成本模板形态并生成 adoption proof

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

command-checkpoint

Codex adapter for Claude slash command /checkpoint - 落盘当前 session 状态到 memory/current-status.md(compact/clear/handoff 前)

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

command-experiment-watch

Codex adapter for Claude slash command /experiment-watch - 只读监控一个长实验(不 kill/restart/改 config)

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

command-feature-split

Codex adapter for Claude slash command /feature-split - 把一个任务拆成��清晰文件所有权的并行 worker packet

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

command-paper-reproduce

Codex adapter for Claude slash command /paper-reproduce - 把 paper claim 映射到代码并找最小可复现路径

0 Updated 1 weeks ago
a-green-hand-jack
Code & Development Listed

command-pr-review

Codex adapter for Claude slash command /pr-review - 以 fresh reviewer 审查当前 diff(correctness/回归/测试/复现/数据安全)

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

command-result-promote

Codex adapter for Claude slash command /result-promote - 把一个实验结果升级为 evidence / paper claim(走证据门槛 + fresh verifier)

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

command-weekly-maintenance

Codex adapter for Claude slash command /weekly-maintenance - 走一遍每周维护清单:跑治理门禁 + 检查 memory/worktree/artifact/hook 状态

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

interactive-plan-doc

当一个阶段需要在动手前与人类对齐 scope 时,用来写中文 plan doc、收人类批注、收敛并做 plan revision commit(human gate)。

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

session-boundary-control

在阶段切换、context 占用过高、或任务树分叉时,决定 continue/compact/clear/branch/fresh-reviewer 并维护 session 树。

0 Updated 1 weeks ago
a-green-hand-jack
Data & Documents Listed

template-feedback

从当前下游 ml-research-project 把「模板需要新增/修改/增强的能力」自动打包成上游 template repo 的 issue;带上模板版本、涉及路径与复现,走 from-downstream 标签。human 不必手写 issue。

0 Updated 1 weeks ago
a-green-hand-jack
AI & Automation Listed

workflow-recipe-harvesting

当要从 human-cc 会话 trace 提炼可复用的 Claude Code 工作流 recipe 时,用来产出 recipe、绑复测任务与报告,并维护 practices 索引。

0 Updated 1 weeks ago
a-green-hand-jack

Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.