Community posts

Claude Code 2.1.265 - Claude DesktopとCoworkがClaude apps gateway経由で送信するテレメトリに、ターミナルセッションと同様の`http://user.email`と`user.groups`を追加。 - `--plugin-dir`をプラグイン群の入ったフォルダに向けられるように。マニフェストを持つ各子フォルダが読み込まれ、実行中に追加・削除された子フォルダも検知される。 - ディスクに保存されるツール結果に1GBの上限を追加。会話内プレビューは、保存されたファイルが切り詰められた場合にその旨を表示する。 - `/model opusplan[1m]`が"Model not found"で拒否されていた不具合を修正。 - managed settingsがスキルのみをプラグインにロックしている場合に`/add-dir <subdirectory>`がサブディレクトリのエージェント読み込みを拒否し、エージェントのみがロックされている場合はエージェントを約束してしまっていた不具合を修正。 - Remote Controlからの`/clear`が、完了前にSessionStartフックと開いているターミナルダイアログを待ってしまっていた不具合を修正。 - `/workflows`のエージェント詳細を改善: ツール呼び出しがrunning・failed・doneのいずれかとしてマークされ、サブエージェントがタスクリストを持つ場合はそれが表示され、Enterで列挙された呼び出しをその入力と結果とともに展開できるように - プロンプト途中で入力するスラッシュコマンドを改善: 単一の候補ではなく一覧としてマッチが表示されるようになり（Tabでフルスクリーン外に開く）、プラグインスキルがベア名で見つかるように - 多くのファイルを読み込む長いセッションの再開時間を改善。 - サイズ上限を超えた画像がデコードできない場合のエラー表示を改善: 上限を示すだけでなく、原因と対処方法を示すようになった。 - `.claude`フォルダのパーミッションオプションを更新し、実際に許可する内容（セッション中、プロジェクトの`.claude`フォルダ（または`~/.claude`）内のファイル編集）を明示するように - [VSCode] 一定期間非アクティブなセッションの自動アーカイブを追加（新設定"Archive inactive sessions"、デフォルト14日）。 https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md#21265

Claude Code Templates now supports Function Hooks The latest version is live on npm, letting you install Function Hooks directly into Claude Code If you don’t know what they are yet, stay tuned. I’ll be sharing examples throughout the week They literally let you hack Claude Code https://github.com/davila7/claude-code-templates/releases

【速報】 Claude Codeを、自分の関数で丸ごと書き換えられる。 提案の名前はFunction Hooks。ただし、まだ実装されていません。 https://x.com/ClaudeDevs/status/2095572891941351550/video/1 GitHubのissueで「これ欲しい？」と賛否を集めている段階です。 何が変わるかというと ・今の拡張は、決められたタイミングにシェルの命令を差し込むだけ ・新しい案は処理そのものを横取りして、次の処理へ渡せる ・繋がり方はExpressやKoaと同じで、登録した順に並ぶ ・画面描画もファイル操作も通信も$という1つの窓口を通るので、裏で何をしたか全部たどれる つまり、拡張が「隙間に差し込む道具」から「本体そのものを書き換える権利」に変わります。 【公式】リポのissueに、アーキテクチャのPDFと動くデモ動画が9本ついてる。 正直これ、通ったらプラグインの作り方ごと変わると思う。 AIにどこまで主導権を渡すかを自分で決める設計は、原則の側から読むと腹落ちが速い。 この下の記事を読むと理解が一気に深まる。マジでおすすめ👇 オープンソースはこちら👇 https://github.com/anthropics/claude-code

Claude Code 2.1.265 has been released. 50 CLI changes Highlights: • Telemetry now sends http://user.email/user.groups from Desktop/Cowork via apps gateway, matching terminal sessions • CLI: --plugin-dir can point to a plugins folder; child manifests auto-load and changes apply live • Tool results saved to disk capped at 1 GB; conversation preview shows when a saved file was truncated Full details are in thread ↓

Solana transaction v1 lands on mainnet this week. Your program is about to STOP WORKING. Changes are mandatory. If your code reads, indexes, sponsors or introspects transactions, something in it is about to break, and the worst failures are silent. I just shipped auditor-skill v7.3 to solanabr @SuperteamBR official repo. So an AI agent can check your repo for every one of them. Here is what changed and what to look for. v0 → v1, the short version: • Max size 1,232 → 4,096 bytes • Compute limit, data-size limit, heap and priority fee move out of ComputeBudget instructions into the message config • Unset limits default to ZERO, not 200k CU. Omit them and the tx fails • Priority fee changes units: v0 was micro-lamports per CU, v1 is a total in lamports • ComputeBudget instructions still execute, as no-ops that set nothing • Lookup tables gone, 64 inline accounts, duplicates rejected, max 64 instructions • Base58 can't carry it. Use base64 Sending v1 is opt-in. Reading it is not. Once one v1 tx is in a block: Loud: getTransaction without maxSupportedTransactionVersion: 1 errors. getBlock fails for the entire block. Wedge: blockSubscribe emits block: null and stops advancing. Forever. Silent: every indexer that finds the priority fee by scanning ComputeBudget instructions records zero. No error. Geyser plugins before 15.1.1 downgrade v1 to v0 on the wire. Old protobuf stubs drop the config field. Dangerous: a fee sponsor or paymaster that caps fees by scanning instructions binds nothing on v1. The attacker puts the fee in the config, adds harmless no-op ComputeBudget instructions to keep your scanner happy, and you sign. Repeat until the key is empty. On-chain: a program that reads ComputeBudget instructions from the instructions sysvar to enforce a fee floor is now enforcing nothing. No sysvar or syscall exposes the v1 config. What v7.3 adds: • A vuln-class playbook: fact sheet, who-breaks matrix, per-role worksheets for readers, indexers, sponsors, programs, senders, wallets • KV-135: sponsor cap bypass and disabled ComputeBudget gates • KV-136: reader wedge and zero-budget indexing • 23 checklist items across TypeScript, backend, off-chain Rust, on-chain, deployment, testing, monitoring How to use it in five minutes, two ways: Plugin (Claude Code): 1 - /plugin marketplace add solanabr/auditor-skill 2 - /plugin install auditor 3 - /auditor:audit and tell it "scope: transaction v1 readiness" No install, any agent (Cursor, Codex, Windsurf, Claude Code): 1 - git clone the repo into your project, e.g. .claude/skills/auditor-skill 2 - Tell your agent: "Audit this repo using auditor-skill, scope: transaction v1 readiness. Load references/vuln-classes/transaction-v1.md and known-vectors 135 and 136." 3 - Read the verdicts. Every item is PASS / FAIL / N/A with the file and line It is a folder of markdown. Nothing runs, nothing phones home, delete it when you are done..

10 AI SKILLS AND HARNESS REPOS YOU SHOULD HAVE ON YOUR RADAR If you are building an AI team desk, these 10 open-source repos are worth checking out. 1. skills — Official Anthropic skills catalog for Claude https://github.com/anthropics/skills 2. academic-research-skills — Research → write → review skills for Claude Code https://github.com/Imbad0202/academic-research-skills 3. compound-engineering-plugin — Every compound-engineering plugin for Claude/Codex/Cursor https://github.com/EveryInc/compound-engineering-plugin 4. ponytail — Skill pack still dominating weekly skills chatter https://github.com/DietrichGebert/ponytail 5. gstack — Garry Tan Claude Code setup: CEO/CTO agent tools https://github.com/garrytan/gstack 6. skills — Official Skills Catalog for OpenAI Codex https://github.com/openai/skills 7. marketingskills — CRO, SEO, copy, and growth skills for agents https://github.com/coreyhaines31/marketingskills 8. patent-disclosure-skill — Patent mining / disclosure skill for Claude https://github.com/handsomestWei/patent-disclosure-skill 9. everything-claude-code — One-stop Claude Code tips, skills, and workflows https://github.com/WorldFlowAI/everything-claude-code 10. archify — Claude skill for verifiable architecture diagrams https://github.com/tt-a1i/archify All open source. All worth exploring. Bookmark this for your next AI project.

GROK BUILD HOW TO USE Sources: official http://docs.x.ai/build + http://x.ai/cli + your Grok Build bookmark folder (286 posts) Updated: 2026-09-08 WHAT NEW TODAY - Folder now 286 posts (2 new vs 284 on Sep 7) - Testers (cb_doge, Sep 7): Blender rocket-launch-site from scratch with Grok Build — handles hard parts, hours saved, speed and quality called out as a creative/3D workflow win - Testers (cb_doge, Sep 7): also covered v1.0.22 — smoother workflows, stronger desktop/MCP, clearer file-edit previews, redesigned dashboard, better safety/reliability WHAT IT IS - A terminal-native coding agent with a fullscreen TUI - Also runs headlessly in scripts, or inside other apps via ACP - Understands the repo, edits files, runs shell commands, and manages long jobs - Open source harness: - Different from Grok Bot (cloud teammates) and Digital Optimus (real-time video control) - Bookmark testers: native /remote-control support is coming (SpaceXAI confirmed) INSTALL AND UPDATE [] macOS / Linux / WSL: curl -fsSL | bash [] Windows PowerShell: irm | iex [] Then: grok --version [] Update: grok update (stable) or grok update --alpha - Testers: v1.0.5 adds GROK_CONFIG env, auto worktree reclaim, hook-block wording, RTL, ACP effort, image/video call limits - Testers: v1.0.6 changes subagent spawning (breaking) and adds a customizable status line - Testers: v1.0.7 adds a first-class workflows catalog, Always/Never rules, and a timer-refresh status line - Testers: v1.0.8 improves MCP servers and workflows. Concurrent subagents start faster without freezing the parent or UI. Follow-ups send while a task is running. Ctrl+S stashes the prompt draft. /workflow autocompletes. MCP can request form input or URL consent. Workflow rows show current context usage - Testers: v1.0.9 adds agent budget + reasoning controls, plugin-provided agents, /feedback images, instant TUI mode switching, smarter clone/worktrees, resilient subagents, faster MCP startup, concurrent agents without rate-limit collapse - Testers: v1.0.10 — grok clone reuses matching local checkouts as linked worktrees for faster session creation - Testers: v1.0.11 — Auto/headless QoL: resume-picker headless sessions, configurable default permission mode, Auto subagent messages, mkdir/touch without prompts, no 10-hour monitor timeout, faster background waits, turn duration after /resume - Testers: v1.0.12 — faster worktrees; MCP retry; subagent wait after interjection; hook descriptions; context/token accuracy; auto-compaction bar; .NET watcher fix; Auto recap during background streams - Testers/Official: v1.0.13 · Aug 28 — compressed CLI downloads; truncated responses continue; hooks can confirm (not only allow/deny); inference retry on stalls/drops/5xx; Windows ~/.grok and worktrees; more reliable session save; faster MCP startup when auth configured; truncated tool calls with complete args execute; large-image session fix; scheduled-task UUID IDs + stop reminder - Testers: v1.0.14 — PostToolUse hooks feed model feedback; grok usage <session-id> per-turn tokens/cost; proactive OIDC refresh; stronger subagents; ~70% smaller Windows CLI downloads; large-session memory no longer blocks turns - Testers: v1.0.15 — background model connect/token refresh/MCP at startup (faster first reply + new sessions); memory consolidation deferred off exit path; /copy and /export tip after repeated drag-copies - Testers: v1.0.16 — MCP servers at session bind for workspace integrations; enterprise model restrict via signed requirements.toml; 1h default wait ceiling for long subagent/task output; MCP OAuth from /mcps deadlock fix; sessions recover after expired-token network interrupts - Testers: v1.0.17 — MCP multi-round-trip elicitation (input_required → resume); quieter next-prompt ghost suggestions; /btw panel table cell select/copy as TSV - Testers: v1.0.18 — managed policy blocks disallowed MCP servers and marketplace installs before write; interactive Chart.js charts; richer image-gen params via composer state; configurable per-model rate-limit retry; per-model mTLS for secure upstream; faster startup/first responses with more work in background; session/subagent/auth/composer/UI fixes - Testers: v1.0.19 — scheduled /loop tasks always run in background (no turn inject); session resume reports running loops/subagents/workflows; headless `grok -p --worktree`; dashboard shows new sessions immediately + /usage from dashboard; transparent terminal theme; faster first prompt after login with MCP; UI freeze/crash fixes - Testers: v1.0.22 — first-party desktop MCP; `grok-workspaced daemon` for Computer Hub folder expose; continue finished subagents; background agents return real output; real line-number diffs + auto-expand on edit permission; built-in tools beat user MCP name collisions; remote-control device grouping; Auto blocks destructive `git checkout --`; faster open/resume (MCP config off critical path) - Bookmark testers: Grok Build mode is now on all SuperGrok and X Premium plans, web and mobile - Testers: official Netlify plugin for build and deploy - Testers: Quo communications plugin — SMS (individual/group/bulk), contacts, tasks, message/call transcripts, missed-call follow-up (`grok plugin install quo --trust`) - Bookmark testers: Grok Build can find/sequence official footage and produce production-quality video edits from a brief (or your own clips) - Bookmark testers: Grok Build Mode can ship simple web apps (example: image watermarker) - Bookmark testers: Grok Build can drive Blender 3D workflows (example: rocket launch site built from scratch) - Bookmark testers: a usage-reset control can set usage back to 0% - Changelog: - Evaluate Grok 4.6 with the Grok Build harness (http://grok.com/build) - Docs: - Download page: SIGN IN [] First launch opens a browser for authentication [] Headless or remote: grok login --device-auth [] Or: export XAI_API_KEY="xai-..." then grok [] Sign out: grok logout [] In TUI: /login and /logout START A SESSION [] cd your-project then grok [] First prompts: Explain this repo. or src/main.rs Walk me through this file. [] grok -p "Explain this codebase" for a one-shot headless run [] grok -c continues the most recent session in this directory [] grok -r resumes a session by id (or the most recent) [] grok inspect shows discovered config, skills, plugins, hooks, and MCP [] grok models lists available models [] Switch model in TUI with /model LEARN THE TOOL [] Run /tutorial or /help [] /release-notes or /changelog for the current version [] Read AGENTS.md in the repo (also CLAUDE.md if you are migrating) [] Import Claude settings with /import-claude [] Carry over rules, skills, and MCP from Claude / Cursor / Codex [] Keyboard shortcuts: /help and Ctrl+. [] Voice: Ctrl+Space or /voice [] Shift+Tab cycles Plan, Auto, Always-approve HOW TO WRITE A TASK [] Outcome: what should be finished [] Sources: files, folders, URLs, or mentions [] Constraints: what not to touch, and when to ask [] Deliverable: diff, report, test, or published app [] Review point: plan first, or stop before commit/push [] For big work: /plan first, then approve [] For long ownership: /goal [] For research first: /deep-research [] For a sick laptop: tell it what is broken and let it investigate [] Testers (tetsuo): start with work you can check — name the behavior, give a reproduction command, limit files, list final checks, say where to stop [] Testers (tetsuo): inspect first (git status --short, grok inspect); give executable feedback (compiler/linter/tests) so the agent can loop; review the full diff and the command behind every green check CORE COMMANDS [] /new starts a new session [] /home returns to the welcome screen [] /resume and /sessions switch sessions [] /fork branches the current session [] /rename or /title names the session [] /session-info shows session details (click or drag to copy) [] /context shows token and skill/MCP cost [] /usage shows credits [] /compact condenses history [] /rewind truncates conversation history (confirms; does not wipe files) [] /export saves the conversation [] /btw asks a side question without derailing the turn [] /check-work has a subagent review the code [] /feedback files a report [] /doctor and grok doctor fix terminal, tmux, clipboard, keyboard [] grok du shows disk use under ~/.grok [] grok wrap runs a command with local clipboard support [] grok trace exports debug traces [] /loop runs a prompt on an interval (auto-expires after 7 days) [] Queued follow-ups can send as interjections if you turn that setting on [] Ctrl+4 opens the Agent Dashboard [] Ctrl+U can continue a Claude Code session inside Grok Build PLAN MODE AND PERMISSIONS [] /plan or Shift+Tab enters plan mode [] Only the plan file can be edited until you approve [] Auto and always-approve do not skip plan review [] a approves, s requests changes, c comments, q quits plan [] /view-plan reopens the preview [] /auto auto-approves safer tools [] /always-approve skips most prompts (deny rules and hooks still apply) [] Edit always-allow bash as free-form globs [] Restrict web search domains with [toolset.web_search] in config.toml [] PreToolUse hooks can rewrite tool input, not only allow or deny [] StopCancelled reports when a turn is interrupted or rejected [] Pin auth to API key or OIDC in config.toml if needed PARALLEL WORK AND AGENTS [] Agent Dashboard: /dashboard or Ctrl+4 [] Run feature, review, fix, and test agents from one place [] Pause, check progress, and jump into any agent [] Worktrees: grok -w or Ctrl+W (isolated git copy under ~/.grok/worktrees) [] grok worktree list / show / rm / gc [] /fork --worktree forks into a worktree [] /goal owns a long engineering objective (status, pause, resume, clear) [] /create-workflow authors a repeatable multi-agent pipeline [] /workflow launches, pauses, resumes, or stops a run [] /workflows is the live run dashboard [] /deep-research starts a built-in research workflow [] Tools and MCP servers receive GROK_SESSION_ID [] Background tasks and TODOs survive compaction VOICE AND MEDIA [] Dictate with Ctrl+Space [] /imagine generates an image [] /imagine-video generates a video [] Ask it to boost, compress, and re-encode video audio in plain English [] One-prompt 3D: ask it to model in Blender and watch it drive the app [] Morning voice-memo automations can research and read you a briefing [] Disable image/video tools via config or env if you do not want them SKILLS PLUGINS AND MCP [] /marketplace installs plugins [] /plugins /skills /hooks /mcps open the same extensions modal [] Skills live in ./.grok/skills, ~/.grok/skills, and plugin skill folders [] User skills appear as slash commands; use /local:name if names collide [] grok plugin install / enable / disable [] grok mcp list / add / doctor [] Type / to reference a skill [] Common marketplace plugins: Exa, TinyFish, MongoDB, Vercel, Sentry, Cloudflare, Chrome DevTools, Supabase, Neon, Firecrawl, Figma, Railway, Stripe [] Official/Testers: browser-use plugin — /plugins or `grok plugin install browser-use --trust`; enable Chrome remote debugging to drive your signed-in local Chrome (or isolated cloud browser; local uvx, no API key for local Chrome) [] Managed MCP servers come from the gateway catalog [] Add a local directory as a marketplace source if you keep private plugins WORKTREES AND FILES [] Worktrees start from current HEAD, including uncommitted changes [] grok -w --ref main starts from a clean ref [] Ending a session does not delete the worktree; run grok worktree gc [] Keep durable files in the repo or ~/.grok, not temp dirs [] Config: ~/.grok/config.toml (Windows: %USERPROFILE%\.grok\config.toml) BUILD MODE AND PUBLISHING [] Build Mode on http://grok.com (and mobile where available) [] One prompt can become an app on a http://grok.me domain [] Export source to GitHub [] Examples: http://driver.grok.me, http://bbox.grok.me, http://gravity.grok.me [] SuperGrok Heavy may be required for some publish features [] Bookmark testers: Grok Build is now available for free users too — describe what you want, Grok builds apps/websites/games/dashboards, then publish and share [] Testers: Tesla Owners Silicon Valley — creating from a prompt is open to free users; a live http://grok.me URL still requires a paid plan [] Testers: Android (X Freeze) — GitHub push, secrets store, invite-only lock, download builds, custom domains, share to X; end-to-end ship from a phone HEADLESS AND AUTOMATION [] grok -p "task" for scripts [] grok -p "task" --output-format streaming-json [] grok agent stdio for ACP in other apps [] grok serve --protocol acp --port 9120 for a local ACP server [] Custom models go in ~/.grok/config.toml under [http://model.name] COMMUNITY TOOLS [] GrokTerm: multi-harness terminal host with voice [] desktop-harness: Mac Accessibility control (http://github.com/xfreeze2/desktop-harness) [] Quill: universal SuperGrok voice layer on macOS [] AFK Pilot / Grok Build Desktop: remote edit and syntax highlight [] Grok Build for VS Code (Community) extension — Testers: Routines dashboard for scheduled executions across Grok Build, Codex, and Claude Code (89K+ installs; not affiliated with SpaceXAI) [] Unity CLI / MCP for game work SAFETY AND USAGE [] Prefer plan mode before large edits [] Prefer always-approve only after you trust sandbox and deny rules [] Opt in to share traces only if you want to (off by default) [] Privacy banner can be dismissed from Settings [] Video tools may require ZDR storage [] Check /usage before long multi-agent days [] Grok 4.6 may run double-token promos; check current offers OFFICIAL LINKS [] Overview: [] Modes and commands: [] Plan mode: [] Worktrees: [] Skills and plugins: [] CLI reference: [] Changelog: [] Install: [] Source:

用 Claude Code、Cursor 或者 Codex 写大工程，最怕上下文滚着滚着就失控，写到后面全是 bug 还巨费 Token。 今天冲上 GitHub Trending 的 ECC 就是干这个的。它不是个简单的 Prompt 词包，而是给智能体加了一整套工程化治理框架：260+ 模块化技能、60+ 专注子 Agent，外加一套自动化 TDD 测试验收闭环。Agent 必须先梳理验收边界并跑通测试才能交差。 * 核心定位：编码 Agent 运行框架与调度增强系统 * 开源协议：MIT License * 兼容平台：Claude Code、Codex CLI、Cursor、OpenCode、Zed * 安装方式：全局运行 npx ecc-universal setup，或在 Claude Code 内执行 /plugin install ecc@ecc * 亮点特性：Session 级上下文剪枝与记忆存储，内置测试回环大幅压低废 Token 浪费 https://github.com/affaan-m/ECC https://ecc.tools/

STOP WASTING 90% OF YOUR CLAUDE CODE TOKENS ON FILE READING Spotify open sourced a plugin that makes your Claude allowance last roughly 10x longer on large file tasks Shunt catches full reads above 350 lines and sends them through Portal to a cheaper worker model such as Gemini 2.5 Flash. Claude gets the answer instead of eating thousands of lines for breakfast It also delegates predictable work: > Tests > Configs > Type stubs > Docs Claude stays on the expensive stuff: debugging, precise edits and architecture The difference is slightly stupid On Spotify’s 162k line Java monorepo, reading one source and test pair dropped from 75,990 Claude tokens to 4,148. Another task fell from 16,221 to 821 That’s 94% fewer Claude tokens while keeping Claude in charge of the decisions that need actual reasoning Install: > claude plugin marketplace add spotify/portal-ai-plugins > claude plugin install portal@portal > claude plugin install shunt@portal Start a new session and run `/portal:setup` This doesn’t create infinite tokens. It just stops paying frontier model prices for `cat file` https://github.com/spotify/portal-ai-plugins

9 步把 FreeCAD 接到 Claude：一句话"设计个法兰"，它就建模、出图，还能跑有限元应力分析。 《FreeCAD MCP：让本地模型生成可 3D 打印的实体》 --- # FreeCAD MCP server **FreeCAD MCP server** 让你从 Claude Desktop（或任何 MCP 客户端）控制 FreeCAD——设计法兰、玩具车、从 2D 图纸建零件，全程对话式。 ## 9 步设置 1. 克隆仓库：`git clone https://github.com/neka-nat/freecad-mcp.git` 2. 复制 `addon/FreeCADMCP` 到 FreeCAD 插件目录（Windows: `%APPDATA%\FreeCAD\Mod\`；macOS: `~/Library/Application Support/FreeCAD/v1-1/Mod/`；Linux 各发行版见 README） 3. 重启 FreeCAD 4. 工作台列表选 **MCP Addon** 5. 免费CAD MCP 工具栏点 **Start RPC Server** 6. （可选）勾 **Auto-Start Server** 自动启动 7. 安装 `uvx`（Astral 工具） 8. 编辑 Claude Desktop 配置 `claude_desktop_config.json`： ```json { "mcpServers": { "freecad": { "command": "uvx", "args": ["freecad-mcp"] } } } ``` 9. 重启 Claude Desktop——对话里让它"设计一个法兰"就行 ## 工具集（14 个） create_document / create_object / edit_object / delete_object / **execute_code**（任意 Python）/ insert_part_from_library / get_view（截图）/ get_object / **run_fem_analysis**（跑 CalculiX 求解器：最大 von Mises 应力、位移）…… **省 token**：`--only-text-feedback` 只要文本反馈；截图可按调用控制（`include_screenshot` + `view_name`）。 ## 演示（README 有 gif） 法兰、玩具车、从 2D 图纸建零件——都有预览。 原文：https://github.com/neka-nat/freecad-mcp #FreeCAD #MCP #3D打印 #本地AI

Rapaz… hoje o Claude Code ainda é hook de shell. Function Hooks (proposta) quer TypeScript estilo Express/Koa pra mudar o agent por dentro — até o render. Comunidade decide se shippa. https://github.com/anthropics/claude-code/issues/91870

すごい、Anthropic公式の Agent Skills repoが出てた。 Claude Codeならまずこれ。 /plugin marketplace add anthropics/skills 中身は19 skills。 pdf/pptx/xlsx/docxと mcp-builder、template、 specまで読める。 自社ルールをskill化する 雛形として保存したい。 https://github.com/anthropics/skills

10 CODING-AGENT AND IDE REPOS YOU SHOULD HAVE ON YOUR RADAR If you are picking a coding agent or AI IDE, these 10 open-source repos are worth checking out. 1. void — Open-source AI code editor in the Cursor lane https://github.com/voideditor/void 2. Warp — Rust terminal with agentic workflows built in https://github.com/WarpDotDev/Warp 3. gstack — Garry Tan's Claude Code tool pack (CEO / eng / release roles) https://github.com/garrytan/gstack 4. compound-engineering-plugin — Compound Engineering plugin for Claude Code / Codex / Cursor https://github.com/EveryInc/compound-engineering-plugin 5. deer-flow — ByteDance long-horizon SuperAgent harness (research + code + create) https://github.com/bytedance/deer-flow 6. context-mode — Context-window optimizer that sandboxes tool output for coding agents https://github.com/mksglu/context-mode 7. copilot-cli — GitHub Copilot from the terminal https://github.com/github/copilot-cli 8. amazon-q-developer-cli — Amazon Q Developer CLI for agentic coding in the shell https://github.com/aws/amazon-q-developer-cli 9. zed — High-performance collaborative code editor https://github.com/zed-industries/zed 10. kiln — Tooling for building and evaluating AI products / agents https://github.com/kiln-ai/kiln All open source. All worth exploring. Bookmark this for your next AI project.

GROK BUILD HOW TO USE Sources: official http://docs.x.ai/build + http://x.ai/cli + your Grok Build bookmark folder (284 posts) Updated: 2026-09-07 WHAT NEW TODAY - Folder now 284 posts (3 new vs 281 on Sep 6) - Testers (XFreeze / blankspeaker, Sep 7): v1.0.22 — first-party desktop MCP server; `grok-workspaced daemon` exposes folders to Computer Hub; continue finished subagents instead of restarting; background agents return real output; edit diffs show real line numbers + auto-expand on permission; built-in tools beat user MCP name collisions; remote-control pickers group by device; Auto no longer auto-runs destructive `git checkout --`; faster open/resume (MCP config no longer blocks UI) - Testers (nima_owji, Sep 6): built a web app in Grok Build Mode that watermarks images WHAT IT IS - A terminal-native coding agent with a fullscreen TUI - Also runs headlessly in scripts, or inside other apps via ACP - Understands the repo, edits files, runs shell commands, and manages long jobs - Open source harness: - Different from Grok Bot (cloud teammates) and Digital Optimus (real-time video control) - Bookmark testers: native /remote-control support is coming (SpaceXAI confirmed) INSTALL AND UPDATE [] macOS / Linux / WSL: curl -fsSL | bash [] Windows PowerShell: irm | iex [] Then: grok --version [] Update: grok update (stable) or grok update --alpha - Testers: v1.0.5 adds GROK_CONFIG env, auto worktree reclaim, hook-block wording, RTL, ACP effort, image/video call limits - Testers: v1.0.6 changes subagent spawning (breaking) and adds a customizable status line - Testers: v1.0.7 adds a first-class workflows catalog, Always/Never rules, and a timer-refresh status line - Testers: v1.0.8 improves MCP servers and workflows. Concurrent subagents start faster without freezing the parent or UI. Follow-ups send while a task is running. Ctrl+S stashes the prompt draft. /workflow autocompletes. MCP can request form input or URL consent. Workflow rows show current context usage - Testers: v1.0.9 adds agent budget + reasoning controls, plugin-provided agents, /feedback images, instant TUI mode switching, smarter clone/worktrees, resilient subagents, faster MCP startup, concurrent agents without rate-limit collapse - Testers: v1.0.10 — grok clone reuses matching local checkouts as linked worktrees for faster session creation - Testers: v1.0.11 — Auto/headless QoL: resume-picker headless sessions, configurable default permission mode, Auto subagent messages, mkdir/touch without prompts, no 10-hour monitor timeout, faster background waits, turn duration after /resume - Testers: v1.0.12 — faster worktrees; MCP retry; subagent wait after interjection; hook descriptions; context/token accuracy; auto-compaction bar; .NET watcher fix; Auto recap during background streams - Testers/Official: v1.0.13 · Aug 28 — compressed CLI downloads; truncated responses continue; hooks can confirm (not only allow/deny); inference retry on stalls/drops/5xx; Windows ~/.grok and worktrees; more reliable session save; faster MCP startup when auth configured; truncated tool calls with complete args execute; large-image session fix; scheduled-task UUID IDs + stop reminder - Testers: v1.0.14 — PostToolUse hooks feed model feedback; grok usage <session-id> per-turn tokens/cost; proactive OIDC refresh; stronger subagents; ~70% smaller Windows CLI downloads; large-session memory no longer blocks turns - Testers: v1.0.15 — background model connect/token refresh/MCP at startup (faster first reply + new sessions); memory consolidation deferred off exit path; /copy and /export tip after repeated drag-copies - Testers: v1.0.16 — MCP servers at session bind for workspace integrations; enterprise model restrict via signed requirements.toml; 1h default wait ceiling for long subagent/task output; MCP OAuth from /mcps deadlock fix; sessions recover after expired-token network interrupts - Testers: v1.0.17 — MCP multi-round-trip elicitation (input_required → resume); quieter next-prompt ghost suggestions; /btw panel table cell select/copy as TSV - Testers: v1.0.18 — managed policy blocks disallowed MCP servers and marketplace installs before write; interactive Chart.js charts; richer image-gen params via composer state; configurable per-model rate-limit retry; per-model mTLS for secure upstream; faster startup/first responses with more work in background; session/subagent/auth/composer/UI fixes - Testers: v1.0.19 — scheduled /loop tasks always run in background (no turn inject); session resume reports running loops/subagents/workflows; headless `grok -p --worktree`; dashboard shows new sessions immediately + /usage from dashboard; transparent terminal theme; faster first prompt after login with MCP; UI freeze/crash fixes - Testers: v1.0.22 — first-party desktop MCP; `grok-workspaced daemon` for Computer Hub folder expose; continue finished subagents; background agents return real output; real line-number diffs + auto-expand on edit permission; built-in tools beat user MCP name collisions; remote-control device grouping; Auto blocks destructive `git checkout --`; faster open/resume (MCP config off critical path) - Bookmark testers: Grok Build mode is now on all SuperGrok and X Premium plans, web and mobile - Testers: official Netlify plugin for build and deploy - Testers: Quo communications plugin — SMS (individual/group/bulk), contacts, tasks, message/call transcripts, missed-call follow-up (`grok plugin install quo --trust`) - Bookmark testers: Grok Build can find/sequence official footage and produce production-quality video edits from a brief (or your own clips) - Bookmark testers: Grok Build Mode can ship simple web apps (example: image watermarker) - Bookmark testers: a usage-reset control can set usage back to 0% - Changelog: - Evaluate Grok 4.6 with the Grok Build harness (http://grok.com/build) - Docs: - Download page: SIGN IN [] First launch opens a browser for authentication [] Headless or remote: grok login --device-auth [] Or: export XAI_API_KEY="xai-..." then grok [] Sign out: grok logout [] In TUI: /login and /logout START A SESSION [] cd your-project then grok [] First prompts: Explain this repo. or src/main.rs Walk me through this file. [] grok -p "Explain this codebase" for a one-shot headless run [] grok -c continues the most recent session in this directory [] grok -r resumes a session by id (or the most recent) [] grok inspect shows discovered config, skills, plugins, hooks, and MCP [] grok models lists available models [] Switch model in TUI with /model LEARN THE TOOL [] Run /tutorial or /help [] /release-notes or /changelog for the current version [] Read AGENTS.md in the repo (also CLAUDE.md if you are migrating) [] Import Claude settings with /import-claude [] Carry over rules, skills, and MCP from Claude / Cursor / Codex [] Keyboard shortcuts: /help and Ctrl+. [] Voice: Ctrl+Space or /voice [] Shift+Tab cycles Plan, Auto, Always-approve HOW TO WRITE A TASK [] Outcome: what should be finished [] Sources: files, folders, URLs, or mentions [] Constraints: what not to touch, and when to ask [] Deliverable: diff, report, test, or published app [] Review point: plan first, or stop before commit/push [] For big work: /plan first, then approve [] For long ownership: /goal [] For research first: /deep-research [] For a sick laptop: tell it what is broken and let it investigate [] Testers (tetsuo): start with work you can check — name the behavior, give a reproduction command, limit files, list final checks, say where to stop [] Testers (tetsuo): inspect first (git status --short, grok inspect); give executable feedback (compiler/linter/tests) so the agent can loop; review the full diff and the command behind every green check CORE COMMANDS [] /new starts a new session [] /home returns to the welcome screen [] /resume and /sessions switch sessions [] /fork branches the current session [] /rename or /title names the session [] /session-info shows session details (click or drag to copy) [] /context shows token and skill/MCP cost [] /usage shows credits [] /compact condenses history [] /rewind truncates conversation history (confirms; does not wipe files) [] /export saves the conversation [] /btw asks a side question without derailing the turn [] /check-work has a subagent review the code [] /feedback files a report [] /doctor and grok doctor fix terminal, tmux, clipboard, keyboard [] grok du shows disk use under ~/.grok [] grok wrap runs a command with local clipboard support [] grok trace exports debug traces [] /loop runs a prompt on an interval (auto-expires after 7 days) [] Queued follow-ups can send as interjections if you turn that setting on [] Ctrl+4 opens the Agent Dashboard [] Ctrl+U can continue a Claude Code session inside Grok Build PLAN MODE AND PERMISSIONS [] /plan or Shift+Tab enters plan mode [] Only the plan file can be edited until you approve [] Auto and always-approve do not skip plan review [] a approves, s requests changes, c comments, q quits plan [] /view-plan reopens the preview [] /auto auto-approves safer tools [] /always-approve skips most prompts (deny rules and hooks still apply) [] Edit always-allow bash as free-form globs [] Restrict web search domains with [toolset.web_search] in config.toml [] PreToolUse hooks can rewrite tool input, not only allow or deny [] StopCancelled reports when a turn is interrupted or rejected [] Pin auth to API key or OIDC in config.toml if needed PARALLEL WORK AND AGENTS [] Agent Dashboard: /dashboard or Ctrl+4 [] Run feature, review, fix, and test agents from one place [] Pause, check progress, and jump into any agent [] Worktrees: grok -w or Ctrl+W (isolated git copy under ~/.grok/worktrees) [] grok worktree list / show / rm / gc [] /fork --worktree forks into a worktree [] /goal owns a long engineering objective (status, pause, resume, clear) [] /create-workflow authors a repeatable multi-agent pipeline [] /workflow launches, pauses, resumes, or stops a run [] /workflows is the live run dashboard [] /deep-research starts a built-in research workflow [] Tools and MCP servers receive GROK_SESSION_ID [] Background tasks and TODOs survive compaction VOICE AND MEDIA [] Dictate with Ctrl+Space [] /imagine generates an image [] /imagine-video generates a video [] Ask it to boost, compress, and re-encode video audio in plain English [] One-prompt 3D: ask it to model in Blender and watch it drive the app [] Morning voice-memo automations can research and read you a briefing [] Disable image/video tools via config or env if you do not want them SKILLS PLUGINS AND MCP [] /marketplace installs plugins [] /plugins /skills /hooks /mcps open the same extensions modal [] Skills live in ./.grok/skills, ~/.grok/skills, and plugin skill folders [] User skills appear as slash commands; use /local:name if names collide [] grok plugin install / enable / disable [] grok mcp list / add / doctor [] Type / to reference a skill [] Common marketplace plugins: Exa, TinyFish, MongoDB, Vercel, Sentry, Cloudflare, Chrome DevTools, Supabase, Neon, Firecrawl, Figma, Railway, Stripe [] Official/Testers: browser-use plugin — /plugins or `grok plugin install browser-use --trust`; enable Chrome remote debugging to drive your signed-in local Chrome (or isolated cloud browser; local uvx, no API key for local Chrome) [] Managed MCP servers come from the gateway catalog [] Add a local directory as a marketplace source if you keep private plugins WORKTREES AND FILES [] Worktrees start from current HEAD, including uncommitted changes [] grok -w --ref main starts from a clean ref [] Ending a session does not delete the worktree; run grok worktree gc [] Keep durable files in the repo or ~/.grok, not temp dirs [] Config: ~/.grok/config.toml (Windows: %USERPROFILE%\.grok\config.toml) BUILD MODE AND PUBLISHING [] Build Mode on http://grok.com (and mobile where available) [] One prompt can become an app on a http://grok.me domain [] Export source to GitHub [] Examples: http://driver.grok.me, http://bbox.grok.me, http://gravity.grok.me [] SuperGrok Heavy may be required for some publish features [] Bookmark testers: Grok Build is now available for free users too — describe what you want, Grok builds apps/websites/games/dashboards, then publish and share [] Testers: Tesla Owners Silicon Valley — creating from a prompt is open to free users; a live http://grok.me URL still requires a paid plan [] Testers: Android (X Freeze) — GitHub push, secrets store, invite-only lock, download builds, custom domains, share to X; end-to-end ship from a phone HEADLESS AND AUTOMATION [] grok -p "task" for scripts [] grok -p "task" --output-format streaming-json [] grok agent stdio for ACP in other apps [] grok serve --protocol acp --port 9120 for a local ACP server [] Custom models go in ~/.grok/config.toml under [http://model.name] COMMUNITY TOOLS [] GrokTerm: multi-harness terminal host with voice [] desktop-harness: Mac Accessibility control (http://github.com/xfreeze2/desktop-harness) [] Quill: universal SuperGrok voice layer on macOS [] AFK Pilot / Grok Build Desktop: remote edit and syntax highlight [] Grok Build for VS Code (Community) extension — Testers: Routines dashboard for scheduled executions across Grok Build, Codex, and Claude Code (89K+ installs; not affiliated with SpaceXAI) [] Unity CLI / MCP for game work SAFETY AND USAGE [] Prefer plan mode before large edits [] Prefer always-approve only after you trust sandbox and deny rules [] Opt in to share traces only if you want to (off by default) [] Privacy banner can be dismissed from Settings [] Video tools may require ZDR storage [] Check /usage before long multi-agent days [] Grok 4.6 may run double-token promos; check current offers OFFICIAL LINKS [] Overview: [] Modes and commands: [] Plan mode: [] Worktrees: [] Skills and plugins: [] CLI reference: [] Changelog: [] Install: [] Source:

A ```dot fence in Claude Code becomes a drawing, in place, as the reply streams. To install: /plugin marketplace add sureffi/drawer /plugin install drawer@drawer

让 Claude 直接操控 FreeCAD，用自然语言生成可打印的 3D 模型。 neka-nat/freecad-mcp 的初衷：把开源 CAD 软件 FreeCAD 变成 Claude Desktop 的工具，让模型不再只是"描述"几何，而是真正"动手"建零件。 步骤 0：安装 - 安装 FreeCAD（开源，免费） - 安装 uv / uvx - 下载 Claude 桌面版（可选，带视觉能力可看截图纠错） 步骤 1：克隆仓库 git clone https://github.com/neka-nat/freecad-mcp.git cd freecad-mcp 步骤 2：安装 FreeCAD 插件 把 addon/FreeCADMCP 复制到 FreeCAD 的 Mod 目录： - Linux：~/.local/share/FreeCAD/v1-1/Mod/ - macOS：~/Library/Application Support/FreeCAD/v1-1/Mod/ - Windows：%APPDATA%\FreeCAD\Mod\ 步骤 3：启动 RPC 服务 打开 FreeCAD → 切换到 "MCP Addon" 工作台 → 点 "Start RPC Server"（可勾选 Auto-Start） 步骤 4：配置 Claude Desktop 编辑 claude_desktop_config.json： { "mcpServers": { "freecad": { "command": "uvx", "args": ["--directory", "/your/path/to/freecad-mcp", "run", "freecad-mcp"] } } } 重启 Claude Desktop。 步骤 5：开始建模 在 Claude 里直接说： "建一个带中心孔的法兰盘，外径 80mm，内孔 20mm，厚度 10mm，导出 STL。" Claude 会调用 create_document、create_object、execute_code 等工具，在 FreeCAD 里真正生成参数化模型，并截图给你看。 步骤 6：导出打印 模型建好后，让 Claude 导出 .STL 或 .STEP，丢进 3D 打印机，就有实物了。 这才是官方设计的范式：Claude 是客户端，FreeCAD 是引擎，MCP 是桥梁。

Agent 画图最烦的不是不会画。 是一出手就是圆角框、箭头、渐变色，一眼 AI。 最近 GitHub 上 Diagram Design 涨得很快。 今天又涨了 600+ Star。 它专门给 Claude Code、Codex 这些 Agent 补“画图审美”。 目前内置 39 种图： 架构图 Sequence Diagram 流程图 数据流 状态机 ER 图 时间线 Swimlane Gantt Sankey Fishbone Wardley Map UML Class 数据库 Schema User Journey 等等。 生成结果不是 Mermaid。 直接输出独立 HTML + SVG。 最新版还刚补了 Sankey、Fishbone、Kanban、Deployment、Dependency Graph、Story Map 等 10 种布局。 Claude Code 安装： /plugin marketplace add cathrynlavery/diagram-design /plugin install diagram-design@diagram-design Codex、Pi、Factory Droid、Kiro、OpenCode 也能接。 现在已经 32K+ Star。 天天让 Agent 画架构图、流程图、技术文档的，这个 Skill 可以存一下。 https://github.com/cathrynlavery/diagram-design?utm_source=chatgpt.com

diagram-design — skill open-source pra Claude Code/Codex gerar 39 tipos de diagramas editoriais em HTML+SVG. resolve o problema clássico do "AI gera diagramas feios com caixas de canto reto". pesquisa oficial (http://github.com/cathrynlavery/diagram-design): → 39 tipos visuais cobrindo: architecture, IT current-state, flowchart, sequence, state machine, ER/data model, timeline, swimlane, quadrant, radar/spider, loop/flywheel, nested, tree, org chart, layer stack, Venn, pyramid/funnel, bar, line, treemap, Gantt, scatter, high-level, process, medallion, data flow, DP integration, DP security matrix, Sankey, fishbone, Wardley map, kanban, user journey, deployment, dependency graph, UML class, story map, database schema, polar chart → 3 variantes cada tipo: minimal light, minimal dark, full-editorial — você escolhe → saída: HTML + inline SVG, self-contained, sem build step, sem JS, sem dependência externa → brand matching: passa URL do seu site e o skill extrai palette + tipografia automaticamente (~60 segundos) → redraw de http://draw.io ou Mermaid: passa fonte .drawio ou .mmd e ele regenera com a qualidade editorial → export PNG via Playwright (2× default, configurável até 3×) → export SVG standalone (abre em Figma, Illustrator, browser) 7 semantic patterns (comportamento separado do visual): → fan-in queues + bottlenecks → repeated stage slots → unstructured-input transformation → paired policy traces → secure paved roads → governance catalogs → compensating security layers cada pattern define: triggers, primitives, budget, anti-patterns, fallback estático, e tipo visual mais próximo. agent escolhe pattern primeiro, tipo depois. limites por tipo (cap intencional pra legibilidade): → sankey: max 3 estágios / 8 nós / 12 fluxos → fishbone: 6 bones, 3 sub-causas → kanban: 5 colunas / 12 cards → dependency graph: 9 nós / 14 edges / 4 ranks / 1 cycle → UML class: 7 classes / 8 relationships → annotation callouts: 2 (cap em todo diagrama) acessibilidade: → role="img" + aria-labelledby resolvido → IDs prefixados por diagrama+variante (sem colisão se inlinear múltiplos) → decorative icons escondidos de assistive tech → reduced-motion: mostra frame estático completo, sem playback controls instalação (Claude Code): /plugin marketplace add cathrynlavery/diagram-design /plugin install diagram-design@diagram-design export via slash command: /diagram-design:export-diagram path/to/diagram.html /diagram-design:export-diagram path/to/diagram.html --svg-only /diagram-design:export-diagram path/to/diagram.html --png-only --scale=3 pra dev brasileiro: quem trabalha com diagramas em doc técnico, blog post, RFC, ou PRD sabe o quanto "AI gera Mermaid feio" dói. essa skill resolve — diagramas prontos pra publicação sem fazer à mão. #DiagramDesign #ClaudeCode #DiagramSkill #OpenSource #EditorialDesign #SVG #HTML #DevTools #ClaudeCodeSkills #PluginDev #DataViz #DevBR #IndieHacker #BuildInPublic #CodeNewbie #BrazilDev #ShipIt

The Composio plugin for Claude Code can now update itself from its SessionStart hook. Release: composio-plugin-cc v0.2.4 https://github.com/ComposioHQ/composio-plugin-cc/releases/tag/v0.2.4

日常还是用 cc 和 qodercli 多，于是做了个 claude code / qodercli 的自动跑的插件版。每个会话根据对话内容自动起名，挂在 hook 上，不用贴 prompt，也不用确认。 安装是这两行： /plugin marketplace add sorrycc/session-namer /plugin install session-namer@session-namer 原理是两个异步 hook，UserPromptSubmit 和 SessionStart。底层是往 transcript 追加一行 custom-title，和 /rename 是同一个机制。同时做了一些细节优化，比如打招呼、测试之类的消息不起名，等真正开始干活再起。起完名后每 5 轮回看一次，方向变了就升级名字，最多 3 次。 摘要用你写的语言输出，type 保持英文。如果对内置规则不满意，可以在项目里放一个 .claude/session-name.md 自定义，比如 Format: [TICKET] short imperative title in English, max 8 words. Use [NOJIRA] when no ticket is mentioned 。 qodercli 下零配置：自动识别宿主，调 qodercli 而不是 claude，用 Efficient 档模型。 http://github.com/sorrycc/session-namer

sovereign-skills v6.5.11 — 감사 기반 버그수정 릴리즈 정적 리뷰 + 테스트 실행 + 경계입력 재현 + 공식문서 대조. 20개 스킬 전부 최소 1건씩 실제 수정. 배포 차단급 2건: • Codex 설치 안내가 틀려있었음 — OpenAI 공식 스펙상 Codex는 .agents/skills/<name>/SKILL.md를 스캔해서 스킬을 찾지, agents/openai.yaml을 AGENTS.md에 붙이는 방식이 아님. README 5개 언어 + 20개 yaml 파일 전부 수정. • 플러그인 매니페스트 7개가 잘못된 스키마 필드 때문에 claude plugin validate에서 계속 실패 중이었음 — 20개 전부 통과로 수정. 문체 수준이 아니라 실제 재현된 버그들: • pre-push — 시크릿 스캔이 staged diff만 봤는데, git push는 outgoing 커밋도 같이 보냄. 수정. • project-overview — 손상된 마커 쌍이 수동 작성 텍스트를 조용히 삭제할 수 있었음. 인코딩 깨진 파일 하나가 전체 실행을 죽였음. 일부 실패해도 상태는 항상 WORKING. • session-checkpoint — 격리(QUARANTINE)된 항목이 메모리 승격 큐로 새어나갈 수 있었음. attestation 타이밍 문제로 다음 세션에서 거짓 "변조" 경보 발생. • scope — 잘못된 입력(틀린 키, 범위밖 값)이 검증을 조용히 통과. • code-autopsy — 고치기 쉬운 대형사고가 Critical 게이트를 피해갈 수 있었음. 그 외 setup·project-init·project-check·session-start·collab-audit·skill-ops·integration-intake·clean-room·eval-leakage-audit·doc-drift·goal-lock·full-audit(신규 AUDIT_ONLY/PROPOSE/APPLY_APPROVED 모드 게이트)·freeze·next-action·stepback도 수정. 레포 내 테스트 스위트 9개 전부 통과(204개 assertion). 상세는 CHANGELOG.md 참조. git clone https://github.com/AlexZio00/sovereign-skills.git CHANGELOG: https://github.com/AlexZio00/sovereign-skills/blob/main/CHANGELOG.md

ANTHROPIC JUST OPEN SOURCED THE ENTIRE WALL STREET WORKFLOW AND FIRMS ARE NOT GOING TO BE HAPPY ABOUT IT. DCF models. LBO models. Equity research reports. Merger analysis. KYC checks. All of it. Free. On GitHub. Here is what just became available to anyone with a laptop. Direct connections to Bloomberg, FactSet, S&P Global, Morningstar, and PitchBook. Real Excel models with live formulas and sensitivity tables built automatically. CIMs, IC memos, earnings reports, and buyer lists drafted on demand. PE due diligence, GL reconciliation, and NAV tie-outs running as production agents. This is not a chatbot wrapper that summarizes financial news. These are production agents that own entire financial workflows end to end. The kind that investment banks and private equity firms pay $50,000 to $500,000 per year in software licenses to run. Now it is a one-line Claude Code plugin install. 19,800 GitHub stars. Apache 2.0 license. 100% open source. Think about what this actually means. A junior analyst at a bulge bracket bank spends 80% of their 100-hour week running models, drafting memos, and compiling data across Bloomberg and FactSet. That entire workflow just became a Claude Code agent. The banks charging clients $500 an hour for analysis that this system produces in minutes are not going to tell you this exists. The boutique advisory firms charging $50,000 retainers for due diligence work that these agents handle autonomously are not going to promote this repo. But it is already live. 19,800 people have already starred it. The window where knowing this gives you an edge over every analyst, associate, and advisor still doing this manually is open right now. Star it. Fork it. Deploy it this weekend. Bookmark this before your next financial model. Follow @cyrilXBT for every open source release that disrupts an overpriced industry the moment it drops.

Claude Code 经常要查网页、抓数据、做 UI 测试的话，可以装一下 Browserbase Skills。 它把浏览器能力做成了一组 Skill，装好以后直接用自然语言下任务。 比如让 Claude 去 Hacker News 抓评论并总结，或者直接测试本地页面，发现问题以后继续修。 里面还有 Cookie 同步、验证码处理、住宅代理、浏览器 Trace、UI Test 这些能力。 本地 Chrome 的登录状态也能复用，遇到需要登录的网站，不用每次重新走一遍。 Claude Code 里两条命令就能装： /plugin marketplace add browserbase/skills /plugin install browse@browserbase 如果你平时用 Claude Code，又经常要让它操作网页，这套 Skill 可以直接配上。 GitHub：http://github.com/browserbase/skills

Claude Codeの回を重ねると答えが荒くなる人へ。GPL版のスキル集Context Engineering Kitで対策できます。 サブエージェントで13スキルを独立させcontextを守る設計です（1,639スター）。 Reflexion／SDD／DDD／Review／Git／MCP／Docs／Kaizen ほか。 13スキルを1コマンドで既存のClaude Codeに導入できます。 /plugin marketplace add NeoLabHQ/context-engineering-kit 詳細はこちら↓↓

🔌 복잡한 개념을 배경지식 없이 이해할 수 있게 풀어주는 Claude Code 플러그인 ELI5 https://github.com/anthropics/claude-plugins-community/tree/main/eli5 Claude Code 팀의 Thariq가 만들고, Anthropic 내부에서도 자주 쓰는 스킬임. 사실 사용법은 간단함. /eli5 <알고 싶은 주제> 이렇게 입력하면 끝. 결과는 긴 설명이 아니라, 큰 그림과 짧은 문장으로 구성된 HTML 아티팩트. 이 모듈은 어떻게 동작하지?, 왜 이 트레이드오프를 택했지?, 이번 장애의 원인은?.... 이런 것처럼 코드/설계/인시던트를 빠르게 파악할 때 너무 좋음. 커뮤니티 마켓플레이스에서 바로 설치할 수 있는데,, 이미 공개되어 있으니 Codex 등 여기저기서 다 활용할 수 있음. > claude plugin marketplace add anthropics/claude-plugins-community > claude plugin install eli5@claude-community

The WePROXA Plugin is here You can now add it to Codex, Claude Code, VS Code + GitHub Copilot, GitHub Copilot CLI, or Cursor, easily. Codex : codex plugin marketplace add ennbou/weproxa-plugin codex plugin add weproxa@weproxa-plugins Claude Code : claude plugin marketplace add ennbou/weproxa-plugin claude plugin install weproxa@weproxa-plugins repo : https://github.com/ennbou/weproxa-plugin docs : https://weproxa.com/guide/guides/mcp-integration/

One model’s answer can become your engineering blind spot. claude-council is a Claude Code plugin for developers who want to compare several AI coding agents on the same question. It helps you pressure-test architecture decisions, debugging paths, and code reviews by querying configured providers in parallel, showing their answers side by side, and adding a synthesis of agreement and disagreement. Key features: • Multi-provider queries – ask Gemini, OpenAI, Grok, Perplexity, and Kimi at the same time • OpenRouter support – add models available through OpenRouter, with Claude configured as the default route • Local and subscription options – use supported coding CLIs or run a local Ollama model as a council member • Side-by-side streaming – view responses in a tmux side pane as they arrive, with provider and model labels • Decision-depth controls – assign review roles, run a two-round debate, or send longer queries to background jobs It’s open-source (MIT license). Link in the reply 👇

claude code is exploring function hooks custom code at specific points so you can extend behavior without waiting on built-ins demos show injecting checks or transforms mid-run. not shipped yet. feedback open on github. https://x.com/ClaudeDevs/status/2095572891941351550

CLAUDE BUILDS THE SCRAMBLE TEXT EFFECT THE WHOLE INTERNET IS COPYING IN ONE PROMPT that is the scramble text effect the whole internet keeps trying to copy, and Claude builds it for you in one prompt you describe the look and it writes the whole thing each letter flickers through random characters on a fast interval then the real letters lock in one by one from left to right the reveal staggers so the word decodes into place instead of snapping all at once Claude wires it straight to the GSAP ScrambleText plugin that was made for this exact motion then it does the chrome wordmark with background-clip text and a metallic gradient that jumps dark to light silver one sharp white highlight band and flat type reads as polished metal pair a huge bold English wordmark with smaller vertical Chinese characters and you get instant hierarchy and a global editorial feel a designer charges 500 to 2000 for a hero section like this and you get it in an afternoon tomorrow I drop the full prompt that builds this landing page from a blank file

Claude Codeのトークン代を下げてほしい人へ、Spotifyが実測90%減の仕組みを公開しました。 「Portal」プラグインです。pre-tool hookで350行超のファイル読みを止め、workerモデル（Gemini 2.5 Flash等）に回します。 claude plugin marketplace add spotify/portal-ai-plugins /portal:setup Layer 1のhookで読取り前に機械的にブロックが走ります。編集の委譲は不可です。 詳細はこちら↓↓