Mstar Compound logo

Mstar Compound

Community
btspoony
mstar-compound

知识结晶 (compound):把已解决的问题沉淀为 `{KNOWLEDGE_DIR}` 下的结构化知识文档,迭代收口时提升 specs/guides;双轨 Bug/Knowledge,含重叠与可发现性检查。无结晶时以 reasoned `skipped` 结束。

Overview

Publisherbtspoony
Repositorymstar-harness
Skill namemstar-compound
Stars
67
Forks
4
Bundled files
5
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 5 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by btspoony on GitHub. Read the source before you install it.

Installation

Install the Mstar Compound AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/btspoony/mstar-harness/tree/main/skills/mstar-compound
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/btspoony/mstar-harness.git /tmp/mstar-harness
mkdir -p .claude/skills
cp -r /tmp/mstar-harness/skills/mstar-compound .claude/skills/mstar-compound
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Mstar Compound in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Mstar Compound on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Mstar Compound is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

mstar-compound(知识结晶)

Load order

Read mstar-harness-core first. Path symbols ({KNOWLEDGE_DIR}, {HARNESS_DIR}) → mstar-conventions. On conflict, mstar-harness-core wins.

Purpose

After solving a non-trivial problem, mstar-compound captures the learning as a structured document in {KNOWLEDGE_DIR}, so future plan research, debugging, and implementation can find and reuse it.

In the mstar lifecycle, compound is invoked by lifecycle flows — the iteration-close round or a standalone-development delivery tail — sharing one step sequence (see「Integration with mstar lifecycle」). It is never triggered at per-plan Done, and stays invocable standalone for ad-hoc captures.

Knowledge that isn't captured evaporates when the session ends. Knowledge that is captured but not discoverable is equally lost. This skill addresses both.

产物存储位置

SSOT: mstar-conventions/references/artifact-storage-paths.md。本 skill 不重定义路径;知识文档 → {HARNESS_DIR}/knowledge/<category>/<slug>.md,CONCEPTS.md → <repo-root>/CONCEPTS.md。<category> 取值见 references/category-mapping.md。

是否值得结晶(自检门禁)

调用本 skill 前,PM(或触发方)必须对候选问题逐条自检 Q1–Q8(诊断耗时 / 非显而易见性 / 可复用性 / 项目特异性 / 既有重叠 / 架构影响 / 失败尝试价值 / 跨模块)。完整问题表、决策矩阵(Yes≥4 强烈建议;Yes=3 Lightweight;Yes≤2 跳过;Q5=Yes 高重叠不新建)与示例判定 → references/compound-workflow.md「是否值得结晶」。

快判:Q5 高重叠 → 更新已有文档,不新建;其余按决策矩阵 Yes 数。

Integration with mstar lifecycle

Compound 是被生命周期流程调用的知识结晶流程:本 SKILL 只拥有执行步骤(下方 + references/compound-workflow.md);触发时机、cwd、写入面与结果记录面由调用方流程声明。本 SKILL 不复述调用方细节,也不维护调用方清单 —— 调用方各自指向本技能的步骤文件。

iteration-start → [plan lifecycle × N] → iteration-close → mstar-compound → {KNOWLEDGE_DIR} → feeds next iteration's specify/plan。per-plan Done 不是触发点:迭代内 plan 行不各自运行(迭代整体收口时一次),独立交付 plan 走其交付尾段 disposition,不强制新文档(reasoned skipped 有效)。

Iteration package promotion(package 盘点方法)

调用方声明需要盘点迭代 package 时,compound 的盘点方法如下。

路径:{ITERATION_DIR}/<iteration-id>/**(含 guides/、specs/、扁平 .md;默认排除 delivery-compass.md 除非 PM 显式纳入。Legacy 根目录 *-delivery-compass.md 同理排除)。

步骤动作
1. Inventory列出 package 下全部 .md(除默认排除 compass);读各文件 + package README.md(若有)
2. Triage每篇:Promote / Keep snapshot / Skip(理由写入 compound 摘要)
3. Promote值得跨迭代复用 → 走 Q1–Q8(或轻量判定)→ Phase 2 重叠检测 → Phase 3–6 结构化重写进 {KNOWLEDGE_DIR}/(禁止无改写整文件复制)
4. Trace源文件顶栏或 package README:Promoted to: <knowledge-path>;provenance 记在 catalog 关系(mstar catalog link,关系取 derived-from),不在 README 表格加列
5. Summary记录本轮提升篇数、保留快照、跳过及原因(记录面由调用方承载)

Promote 典型:迭代 spec 已验证且指导未来实现;guide 含非显而易见过程知识或失败尝试。Keep snapshot:仅迭代史、已被 {SPECS_DIR}/ 取代的草案、或自检 ≤2 Yes 的琐碎笔记。边界 SSOT → mstar-iteration/references/iteration-artifact-boundaries.md。

When to use / Skip

Use:迭代收口(mstar-iteration § iteration-close)批量回顾;独立触发(非迭代或紧急,任何非平凡问题解决后);重大 bug 修复后(mstar-iteration 未启用时手动);独立 development plan 交付前的 disposition review(PR head 定稿前,见「Integration with mstar lifecycle」)。

Skip:自检 ≤2 Yes;Q5 高重叠(更新已有而非新建);纯机械工作(格式化、依赖升级、typo);问题未经验证。独立交付 plan 的 disposition 义务不因跳过结晶而消失——结晶跳过时以 reasoned skipped 记录结果(记录在 workflow,见「Integration with mstar lifecycle」)。

Two tracks

TrackWhat it capturesSection structure
BugIncident-level fix — "X broke, here's why and how we fixed it"Problem, Symptoms, What Didn't Work, Solution, Why This Works, Prevention
KnowledgeDurable guidance — "this is how we do X, and why"Context, Guidance, Why This Matters, When to Apply, Examples

Track 由 problem_type 决定,完整映射见 references/category-mapping.md。

Execution modes

ModeWhenBehavior
Full (default)Most casesDispatches research subagents for context analysis, solution extraction, overlap detection
LightweightSimple fixes, context-tight sessionsSingle-pass documentation, no subagents, faster

In Cursor, Full mode dispatches subagents via Task tool. PM selects mode.

Workflow skeleton(Phase 1–7)

完整步骤细节(Gather context / Overlap detection / Write document 含 frontmatter schema + path + validate / Discoverability check / CONCEPTS.md synergy / Catalog registration / Refresh trigger)→ references/compound-workflow.md。每 Phase 一个关键决策:

  1. Gather — 读对话史 + iteration package;分类 track/category(references/category-mapping.md)
  2. Overlap — 高重叠 → 更新已有(加 last_updated);中度 → 新建并标 consolidation review;低/无 → 正常新建
  3. Write — path + frontmatter(SSOT references/schema.yaml)+ body(assets/resolution-template.md)+ YAML validate
  4. Discoverability — 若 root AGENTS.md/CLAUDE.md 未提 {KNOWLEDGE_DIR},提议最小补充(需用户同意;拒绝则仅跳过该编辑,doc 仍写)
  5. CONCEPTS.md — 项目特定领域词满足 qualifying bar 时提议入 CONCEPTS.md(规则见 references/concepts-vocabulary.md);全仓 bootstrap 归 mstar-compound-refresh
  6. Catalog — 新文档在 {HARNESS_DIR}/store.db 的 catalog 中登记为 kind=document(document_kind=knowledge):单行 mstar catalog register,或 reviewed mstar catalog discover + mstar catalog import;provenance/supersession 用 mstar catalog link(derived-from / supersedes 关系);生命周期变更用 mstar catalog update(active / archived / superseded;revision 守卫见 help)。不再要求 {KNOWLEDGE_DIR}/README.md 索引行(README 是散文,不是登记表)。知识关联载体是 catalog relations(mstar catalog link);snapshot plans[].metadata.knowledge_refs 仅 legacy 只读,不是写入面。每篇新 doc 必须完成本 Phase
  7. Refresh trigger — 新知识暗示旧 doc 过时 → 推荐 /pm compound-refresh <scope>(不自动跑,仅 flag)

Engine check (when available): run mstar compound validate <doc-path> [--knowledge-dir <dir>] (or import { validateSchemaYaml, assertKnowledgeCatalogCompleteness } from "@mstar-harness/engine" in a host hook) to validate the frontmatter against references/schema.yaml (Phase 3 Write) and assert that every knowledge body under {KNOWLEDGE_DIR} has a catalog row (Phase 6 Catalog, contract §4). The former README index-row assert is retired: the assertIndexRows export refuses actionably (compound.index.retired) instead of passing, and the mstar compound validate --knowledge-dir index-row check fails for the same reason. On fail -> do not proceed; fix and re-run. Skill text below remains authoritative when the runtime is absent.

Support files

读 references/compound-workflow.md(Q1–Q8 + Phase 1–7)、references/schema.yaml(frontmatter)、references/category-mapping.md(track/path 映射)、references/concepts-vocabulary.md(CONCEPTS.md 规则)、assets/resolution-template.md(track 正文结构)。

Skill dependencies

  • mstar-conventions — path symbols({KNOWLEDGE_DIR}、{HARNESS_DIR})
  • mstar-artifacts — store.db catalog 登记与生命周期(snapshot 为迁移源;project register 仅迁移历史)(README index maintenance retired)
  • mstar-compound-refresh — capture 后知识维护;CONCEPTS.md 全仓 bootstrap

NOT to do

  • Do not create a doc for trivial fixes (typos, formatting, dep bumps)
  • Do not write the doc before the solution is verified
  • Do not skip the overlap check — creating duplicates degrades the knowledge store
  • Do not edit AGENTS.md without user consent
  • Do not create CONCEPTS.md entries for general programming vocabulary
  • Do not modify product code — this skill writes documentation only

Evidence

正确结果 = 一篇可发现的结晶文档:{KNOWLEDGE_DIR}/<category>/<slug>.md 通过 references/schema.yaml frontmatter 校验(Phase 3 Write)+ catalog 登记为 kind=document(document_kind=knowledge,Phase 6 Catalog;mstar catalog list 可查到)+ 达标领域词入 CONCEPTS.md(Phase 5)+ 源文件 / package README 标注 Promoted to: <knowledge-path>(Phase 4 Trace)。

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Mstar Compound AI skill do?

知识结晶 (compound):把已解决的问题沉淀为 `{KNOWLEDGE_DIR}` 下的结构化知识文档,迭代收口时提升 specs/guides;双轨 Bug/Knowledge,含重叠与可发现性检查。无结晶时以 reasoned `skipped` 结束。

Why use Mstar Compound on TypingMind?

Because you install it once and use it with any model. Mstar Compound is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Mstar Compound in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/btspoony/mstar-harness/tree/main/skills/mstar-compound. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Mstar Compound?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Mstar Compound?

As many as you like. As long as a model supports skills, you can use Mstar Compound with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Mstar Compound AI skill free?

Yes. It is published on GitHub by btspoony under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇