菜单
Builds and freezes deterministic train, validation, and test corpora with provenance, group-safe splits, manifests, hashes, and prompt-leakage checks. Use before teacher labeling, fine-tuning, or comparing models on generated or retrieved examples.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
基于 SOC 职业分类
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/bastos/skills --skill prepare-training-corpus命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | prepare-training-corpus |
| description | Builds and freezes deterministic train, validation, and test corpora with provenance, group-safe splits, manifests, hashes, and prompt-leakage checks. Use before teacher labeling, fine-tuning, or comparing models on generated or retrieved examples. |
Make the corpus reproducible before any expensive labeling or training call.
Write down:
Separate deterministic scenario generation from teacher labeling. Materialize and validate the entire candidate pool without model calls first.
Do not silently repair generated or teacher-produced rows. Preserve raw failures and derive validated outputs separately.
Use scripts/build_corpus_manifest.py for JSONL split invariants:
python scripts/build_corpus_manifest.py \
--split train=path/train.jsonl \
--split validation=path/valid.jsonl \
--split test=path/test.jsonl \
--id-field id --group-field deckGroupID \
--output path/corpus-manifest.json
Also run project-specific validators for legality, replay, terminal boundaries, data licenses, and context sufficiency. Fail before labeling if:
Freeze exact JSONL bytes, manifests, source revisions, generator configuration, and SHA-256 hashes. Any schema, prompt, or corpus-membership change creates a new revision and fresh state; never resume incompatible evidence.
Writes and revises clear, precise, accessible technical content for any medium. Use for manuals, wikis, README files, developer and user documentation, tutorials, references, specifications, RFCs, feature requests, issues, commit messages, code comments, review comments, pull request messages, release notes, runbooks, decision records, status updates, email, Slack, and other workplace or agent-facing communication.
Audits and designs lean, coherent context across system prompts, persistent instructions, repository guidance, skills, tool descriptions, memory, and references. Use for context engineering; creating, reviewing, simplifying, or reorganizing agent context; resolving conflicting, duplicated, bloated, stale, or misplaced guidance; designing progressive disclosure; or deciding where shared information should live.
Orchestrates and ships a milestone, release, issue set, or multi-workstream engineering goal through one manager and bounded parallel workers. Use when the user asks to manage a milestone, build a dependency graph, execute independent issues in parallel, coordinate threads, sessions, subagents, or worktrees, recover a stopped delivery, or invokes /gogopowermanager.