菜单
Wiki sensemaker. /learn → densify pass. /learn <question> or "drive Q&A" → research-to-thoughts flow. Triggers `learn_pass` MCP tool. Always reads docs first.
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。
基于 SOC 职业分类
Index of available skills — file locations, purpose, when/how to use.
Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.
Architect-mode development for non-trivial features, refactors, and bug fixes. Bundles KISS/DRY/YAGNI principles, phased planning (orient → align → glossary → PRD → TDD → design+delegate), adversarial self-review, and incremental push discipline. Use when starting any non-trivial change, when codebase drifts, or when "AI did the wrong thing" / "code keeps getting worse" symptoms appear. Skip for one-line fixes, throwaway spikes, exploratory reads.
Guide for writing idiomatic Rust code based on Apollo GraphQL's best practices handbook. Use this skill when: (1) writing new Rust code or functions, (2) reviewing or refactoring existing Rust code, (3) deciding between borrowing vs cloning or ownership patterns, (4) implementing error handling with Result types, (5) optimizing Rust code for performance, (6) writing tests or documentation for Rust projects.
| name | learn |
| description | Wiki sensemaker. /learn → densify pass. /learn <question> or "drive Q&A" → research-to-thoughts flow. Triggers `learn_pass` MCP tool. Always reads docs first. |
| tags | ["skill","wiki"] |
First action (mandatory): docs({name: "learn"}) — read fully before any write.
learn_pass({force: true, raise_questions: true})
All multi-doc tools (ingest, mark_question, search, get, update) are batch-only — wrap every payload in {items: [...]}, even for a single record.
search({items: [{query, mode: "qa", include_bodies: true, include_reasons: true, edges_depth: 1}]}) — check suggested_conclusions per result.ingest({items: [
{kind: "thought", body: "[[<question_id>]] ...claim 1..."},
{kind: "thought", body: "[[<question_id>]] ...claim 2..."}
]})
[[<question_id>]] mints a Supports edge — evidence stacks across thoughts.learn_pass({force: true, raise_questions: false}) — promotes a conclusion + hard-deletes the question (lifecycle is open|graveyard|deleted) once support_promote_floor (default 3) Supports edges accumulate or one candidate clears answer_threshold (default 0.6).learn_pass({force:true, support_threshold:0.2, support_promote_floor:1})) or fall back to mark_question({items: [{question_id, status: "deleted"}]}) for already-answered questions, or status: "buried" to park unanswerable ones in questions/graveyard/.