metal
About
The Metal skill analyzes codebases to extract their conceptual architecture as standardized agent, skill, and team definitions. It captures the project's purpose and roles (the "what" and "who") without the implementation details, abstracting them into reusable components. Use it for onboarding, bootstrapping agentic systems, or creating libraries from reference implementations.
Quick Install
Claude Code
Recommendednpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/metalCopy and paste this command in Claude Code to install this skill
Documentation
金
提庫之概念 DNA——其角、程、合作模——成泛 agentskills.io 定。如取貴金於礦,分項目之是(質)與為(施),生可用之技、代理、團定,捕項目之組基因不複其碼。
用
- 入新庫前先繪概念架
- 由既項目啟代理系——將隱流轉為顯之技/代理/團定
- 研項目之組 DNA 以交相作他項目
- 仿參施作技/代理/團庫不襲之
- 解項目之構顯創者之心模與域識
入
- 必:庫或項目根之徑
- 必:旨陳——何為提質?(入、啟、研、交相)
- 可:焦域——項目所注之區(默:全)
- 可:出深——
survey(探+析)、extract(全程)、report(提+書報)(默:extract) - 可:提上限——技+代理+團之總帽(默:15)
礦試
提之中質:
此概念可存於異施乎?
若是——乃金(質)。提之。 若否——乃渣(施末)。棄之。
例:氣象應之概念「合外數源」乃金——施於諸取三方數之項目。「析 OpenWeatherMap v3 JSON 答」乃渣——專於一 API。
提之技述任之類,非具體例。提之代理述角非人。提之團述合作模非組圖。
行
一:探——測礦體
無評探庫構。先繪地後採。
- Glob 樹以解項目之形:
- 源目及其組模(按特、按層、按域)
- 配檔:
package.json、DESCRIPTION、setup.py、Cargo.toml、go.mod、Makefile - 文:
README.md、CLAUDE.md、CONTRIBUTING.md、架構文 - CI/CD:
.github/workflows/、Dockerfile、署配 - 測目及其構
- 讀項目自述(README、包單)以解所稱旨
- 按類/語計檔以衡範與識主技
- 識項目之界——始終、依何、供何
- 成探報:
Project: [name]
Declared Purpose: [from README/manifest]
Languages: [primary, secondary]
Size: [file count, approx LOC]
Shape: [monorepo/library/app/framework/docs]
External Surface: [CLI/API/UI/library exports/none]
得:實測——何在、多大、自稱為何。未分類評。報如地測非評。
敗:無 README 或單→由目名、檔容、測述推旨。項目過大(>1000 源檔)→縮至最活目(用 git log 頻或 README 引)。
二:析——分析成分
讀代表檔以解項目於概念之為。
- 取 5-10 代表檔於不同區——非盡而多元:
- 入點(主檔、路由、命令)
- 核心邏(最引或最被引之模)
- 測(顯欲行勝施)
- 配(顯運憂與署境)
- 各取區,識:
- 域:項目觸何題區?(如「認證」「數轉」「報」)
- 動:項目行何?(如「驗」「轉」「署」「告」)
- 角:碼服何人或系?(如「數工」「終用」「審」)
- 流:何序成流?(如「攝→驗→轉→存」)
- 各見類為:
- 本:諸解此題之施皆有
- 偶:專於本施技之擇
- 成析報:域、動、角、流之表含本/偶標
得:項目之概念圖如域辭典,非碼漫遊。不知技堆者讀之亦解項目之為。
敗:碼隱(重元編、生碼、混淆)→倚測與文非源碼。無測→讀提交以見意。
三:禪——釋施偏
停以清讀碼之認錨。
- 識何架、語、模主吾心模——標之
- 釋於何為:「用 React」變「有 UI 層」。「用 PostgreSQL」變「有持久結構存」。
- 各析報之見施礦試:
- 「合外數源」——可存任處乎?是→金
- 「配 Axios 攔」——可存任處乎?否→渣
- 礦試敗者→更高抽象重書
- 多視角助→以諸鏡視項目:
- 考古:碼構顯創者之心模何?
- 生物:可複基因 vs 具表?
- 樂理:式(奏鳴、輪旋)vs 具音?
- 製圖:何抽象捕有用之拓?
得:析報今無架語。諸見過礦試。概念可移——可施於他語他架之項目。
敗:偏存(見常引具技)→反問:「若此項目重書於異堆,何概念存?」唯彼為金。
四:煉——分金渣
提之核步。各本概念分為技、代理、團。
- 各純析報之本概念,定其類:
Classification Criteria:
+--------+----------------------------+----------------------------+----------------------------+
| Type | What to Look For | Naming Convention | Test Question |
+--------+----------------------------+----------------------------+----------------------------+
| SKILL | Repeatable procedures, | Verb-first kebab-case: | "Could an agent follow |
| | workflows, transformations | validate-input, | this as a step-by-step |
| | with clear inputs/outputs | deploy-artifact | procedure?" |
+--------+----------------------------+----------------------------+----------------------------+
| AGENT | Persistent roles, domain | Noun/role kebab-case: | "Does this require ongoing |
| | expertise, judgment calls, | data-engineer, | context, expertise, or a |
| | communication styles | quality-reviewer | specific communication |
| | | | style?" |
+--------+----------------------------+----------------------------+----------------------------+
| TEAM | Multi-role coordination, | Group descriptor: | "Does this need more than |
| | handoffs, reviews, | pipeline-ops, | one distinct perspective |
| | parallel workstreams | review-board | to accomplish?" |
+--------+----------------------------+----------------------------+----------------------------+
-
各提元:
- 賦泛名——非項目專。「UserAuthService」變
identity-manager(代理)。「deployToAWS()」變deploy-artifact(技)。 - 書一句述不知源亦合
- 注源概念(為跡而非複)
- 末施礦試
- 賦泛名——非項目專。「UserAuthService」變
-
防常分類誤:
- 非各函皆技——尋程非單動
- 非各模皆代理——尋需評斷之角
- 非各合作皆團——尋有專之合作模
- 多項目得 3-8 技、2-4 代理、0-2 團。逾 20→提過細。
得:分類冊各含類(技/代理/團)、泛名、一句述。無項引具技、API、結構。
敗:分類含糊(技乎代理乎?)問:「此關為(技)抑為人(代理)?」技乃方;代理乃廚。仍不明→默為技——技後合易。
五:癒——驗提質
評提誠否——非過非欠。
-
過提察:讀各提定問:
- 可由此復源項目之專邏乎?→過詳
- 引具庫、API、數模、徑乎?→仍渣
- 全施程乎或概念草?→宜草
-
欠提察:唯顯提定(無源項目)問:
- 可解何類項目啟之乎?→宜是
- 定捕項目本性乎?→宜是
- 主能未現乎?→宜否
-
泛察:各定:
- 名於異堆亦合乎?→宜是
- 述無架乎?→宜是
- 可助於異域之項目乎?→理想是
-
衡察:審提比:
- 焦項目典:3-8 技、2-4 代理、0-2 團
- 總少於 3→欠提
- 總多於 15→過提或泛不足
得:信於正抽象層。各定為種可長於異土,非枝唯活原園。
敗:過提→升抽象——合具技為廣,合似代理為一角。欠提→返二取更多檔。泛敗→去技引重書述。
六:鑄——澆金成形
成 agentskills.io 標出文。
- 各提技書骨定:
# Skill: [generalized-name]
name: [generalized-name]
description: [one-line, framework-agnostic]
domain: [closest domain from the 52 existing domains, or suggest a new one]
complexity: [basic/intermediate/advanced]
# Concept-level procedure (3-5 steps, NOT full implementation):
# Step 1: [high-level action]
# Step 2: [high-level action]
# Step 3: [high-level action]
# Derived from: [source concept in original project]
- 各提代理書骨定:
# Agent: [role-name]
name: [role-name]
description: [one-line purpose]
tools: [minimal tool set needed]
skills: [list of extracted skills this agent would carry]
# Derived from: [source role/module in original project]
- 各提團書骨定:
# Team: [group-name]
name: [group-name]
description: [one-line purpose]
lead: [lead agent from extracted agents]
members: [list of member agents]
coordination: [hub-and-spoke/sequential/parallel/adaptive]
# Derived from: [source workflow/process in original project]
- 編諸提入析報——一文含技、代理、團節,加摘表
得:結構報含諸提定於 agentskills.io 式。各定為骨(概念非施),可作 create-skill、create-agent、create-team 之始。
敗:出逾 15→按中性排——留最專於本項目域之概念。常項目皆有之概念(如「manage-configuration」)若無異變則棄。
七:淬——末驗
驗全提且成摘。
- 計提:N 技、N 代理、N 團
- 評覆:跨項目主域乎?
- 驗獨立:讀各定無源項目脈絡——可獨立乎?
- 末施礦試於全集:
Temper Assessment:
+-----+---------------------------+----------+------------------------------------+
| # | Name | Type | Ore Test Result |
+-----+---------------------------+----------+------------------------------------+
| 1 | [name] | skill | PASS / FAIL (reason) |
| 2 | [name] | agent | PASS / FAIL (reason) |
| ... | ... | ... | ... |
+-----+---------------------------+----------+------------------------------------+
- 成末摘:
- 提總(技/代理/團)
- 覆評(何項目域已現)
- 信級(高/中/低)含理
- 建下步:何提定先擴
得:驗析報含摘表、信評、可行下步。報自足——未見源項目者亦可讀解所提概念。
敗:逾兩成提敗末礦試→返四(煉)以更高抽象再提。覆少於識域六成→返二(析)取更多檔。
驗
- 探報含項目構、語、量、所稱旨
- 析識域、動、角、流含本/偶分類
- 禪檢清施偏——出無架語
- 諸提元過礦試(質非施末)
- 技以動名、代理以名名、團以群描述名
- 諸名泛——無項目專引
- 提數於典範(5-15 總非 1 非 30)
- 出定循 agentskills.io 式(前+節)
- 過提與欠提察皆過
- 末淬評含計、覆、信、下步
- 全析報無源亦可解
忌
- 鏡目構:每源檔一技而非提橫切概念。金宜映項目之概念構非檔系。20 檔項目非 20 技
- 拜架:提「configure-nextjs-api-routes」而非「define-api-endpoints」。剝架,留模。礦試捕之:「無 Next.js 可存乎?」若否,乃渣
- 角脹:每模一代理。多項目有 2-5 真角需異專,非 20。尋評斷與溝通異,非僅功能異
- 略礦試:最大敗模。各出必過:「此概念可存於異施乎?」引具庫、API、數模→渣非金
- 生施指:提技宜概念草(3-5 高層步)非全施程。乃由
create-skill養之種,非成品。50 步提乃複非質 - 泛名不足:「UserAuthService」乃類名非概念。「identity-manager」乃角。「manage-user-identity」乃技。由具至共
- 忽合作模:團最難提因合作常隱。尋碼審流、署管、系間數遞、批准鏈——彼顯團構
參
athanorchrysopoeiatransmutecreate-skillcreate-agentcreate-teamobserveanalyze-codebase-for-mcpreview-codebase
GitHub Repository
Related Skills
content-collections
MetaThis skill provides a production-tested setup for Content Collections, a TypeScript-first tool that transforms Markdown/MDX files into type-safe data collections with Zod validation. Use it when building blogs, documentation sites, or content-heavy Vite + React applications to ensure type safety and automatic content validation. It covers everything from Vite plugin configuration and MDX compilation to deployment optimization and schema validation.
polymarket
MetaThis skill enables developers to build applications with the Polymarket prediction markets platform, including API integration for trading and market data. It also provides real-time data streaming via WebSocket to monitor live trades and market activity. Use it for implementing trading strategies or creating tools that process live market updates.
creating-opencode-plugins
MetaThis skill helps developers create OpenCode plugins that hook into 25+ event types like commands, files, and LSP operations. It provides the plugin structure, event API specifications, and implementation patterns for JavaScript/TypeScript modules. Use it when you need to intercept, monitor, or extend the OpenCode AI assistant's lifecycle with custom event-driven logic.
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
