一键在 Manus 中运行任何 Skill

server-start

星标19

分支3

更新时间2026年6月10日 17:54

Start the FOUNDATION dev server (npm run tauri dev) under PM2 management. Use when the app is not running and Claude needs it up (MCP, QA, visual validation) or the user asks to start the server.

安装

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。

在 Manus 中运行

来源

danielterra

danielterra/FOUNDATION

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

Server Start

Starts the dev server as the PM2 process foundation-dev (defined in ecosystem.config.cjs at the repo root).

Steps

Check current state — pm2 describe foundation-dev | Select-String "status" (NOT pm2 jlist | ConvertFrom-Json — PS 5.1 chokes on duplicate env keys). If status is online, report "already running" and STOP — never start a second instance. If stopped/errored, run pm2 delete foundation-dev first. Other PM2 apps (e.g. lightapps) belong to the user — never touch them.
Check for orphans — Get-Process FOUNDATION -ErrorAction SilentlyContinue. An orphan FOUNDATION.exe from a previous unmanaged session blocks the new instance (port/DB). If found, kill it (Stop-Process -Name FOUNDATION -Force) — this is Claude's responsibility now, not the user's.
Start — pm2 start ecosystem.config.cjs (run from repo root, or pass the absolute path).
Wait for readiness — poll the backend log ($env:LOCALAPPDATA\org.w3id.foundation\application.log) for a [DB] Read pool ready or [STARTUP] executor_ready line with timestamp AFTER the start. Poll every ~15s. Budget: up to 5 min when Rust needs recompiling (watch pm2 logs foundation-dev --lines 30 --nostream for cargo build progress), ~60s otherwise.
Report — PID, readiness lines found, and any startup warnings. If the log stalls at MCP server listening with no [STARTUP] lines and pm2 logs shows transport invoke timed out (Vite SSR zombie), STOP and run /server-restart with cache clean instead.

Rules

ALWAYS use PM2 — never spawn npm run tauri dev directly (unmanaged orphan).
NEVER start when status is already online.
The first boot after schema migrations can legitimately take minutes (e.g. one-shot VACUUM) — check the log for [STARTUP] progress lines before declaring a hang.
No console window should ever appear: in debug the exe hides its own orphan console (hide_orphan_console in src-tauri/src/main.rs). A visible blank console window = regression of that guard (and the user closing it kills the server).

同仓库更多 Skills

同仓库

code-review

danielterra/FOUNDATION

Use when reviewing pending code changes (working tree + staged) before commit. Checks against FOUNDATION conventions, architecture layer rules, validates compilation, and flags violations of CLAUDE.md / development.md rules.

2026-06-2419

release-create

danielterra/FOUNDATION

Creates a new FOUNDATION release — bumps version, updates CHANGELOG, SoftwareRelease.ttl, README, commits, tags, and publishes a GitHub release.

2026-06-2419

server-restart

danielterra/FOUNDATION

Restart the FOUNDATION dev server via PM2, with optional Vite/SvelteKit cache clean. Use after config changes, when the Vite SSR module runner goes zombie (transport invoke timed out), or when the user asks for a restart.

2026-06-1019

server-stop

danielterra/FOUNDATION

Stop the FOUNDATION dev server managed by PM2, including orphan child cleanup (FOUNDATION.exe / cargo). Use when the user asks to stop the server or Claude needs a clean shutdown.

2026-06-1019

deps-check

danielterra/FOUNDATION

Use when the user asks to check for available dependency updates in FOUNDATION — phrases like "verifica atualizações de dependências", "o que tem pra atualizar", "tem versão nova das libs", "deps desatualizadas". Runs the Rust deps-check binary (npm + Cargo) and presents the report. Same check that runs automatically on the first session of each day.

2026-06-0819

bug-fix

danielterra/FOUNDATION

Use when the user (PO) asks to fix a Foundation bug by IRI — e.g. "fix bug foundation:Bug_123", "corrija o bug foundation:Bug_456", "resolva o bug <IRI>". Orchestrates the bug pipeline: `support` (investigation & technical dossier) → `architect` Modo Triagem **Briefing** (decide BE/FE/both and produce per-dev briefings) → dispatch `developer-backend` / `developer-frontend` with the briefings → `architect` Modo Triagem **Costura** (validate builds, consolidate `## Como testar`, move to Em Validação (QA)) → `qa` (validation gate before Concluído). Sub-agents cannot call other sub-agents in Claude — this skill (which runs in the main loop) IS the orchestrator. The PO never invokes devs directly; the bug is NEVER closed without QA validation. Always invoke this skill when the user mentions a bug IRI.

2026-06-0619

name	server-start
description	Start the FOUNDATION dev server (npm run tauri dev) under PM2 management. Use when the app is not running and Claude needs it up (MCP, QA, visual validation) or the user asks to start the server.
disable-model-invocation	false

Server Start

Starts the dev server as the PM2 process foundation-dev (defined in ecosystem.config.cjs at the repo root).

Steps

Check current state — pm2 describe foundation-dev | Select-String "status" (NOT pm2 jlist | ConvertFrom-Json — PS 5.1 chokes on duplicate env keys). If status is online, report "already running" and STOP — never start a second instance. If stopped/errored, run pm2 delete foundation-dev first. Other PM2 apps (e.g. lightapps) belong to the user — never touch them.
Check for orphans — Get-Process FOUNDATION -ErrorAction SilentlyContinue. An orphan FOUNDATION.exe from a previous unmanaged session blocks the new instance (port/DB). If found, kill it (Stop-Process -Name FOUNDATION -Force) — this is Claude's responsibility now, not the user's.
Start — pm2 start ecosystem.config.cjs (run from repo root, or pass the absolute path).
Wait for readiness — poll the backend log ($env:LOCALAPPDATA\org.w3id.foundation\application.log) for a [DB] Read pool ready or [STARTUP] executor_ready line with timestamp AFTER the start. Poll every ~15s. Budget: up to 5 min when Rust needs recompiling (watch pm2 logs foundation-dev --lines 30 --nostream for cargo build progress), ~60s otherwise.
Report — PID, readiness lines found, and any startup warnings. If the log stalls at MCP server listening with no [STARTUP] lines and pm2 logs shows transport invoke timed out (Vite SSR zombie), STOP and run /server-restart with cache clean instead.

Rules

ALWAYS use PM2 — never spawn npm run tauri dev directly (unmanaged orphan).
NEVER start when status is already online.
The first boot after schema migrations can legitimately take minutes (e.g. one-shot VACUUM) — check the log for [STARTUP] progress lines before declaring a hang.
No console window should ever appear: in debug the exe hides its own orphan console (hide_orphan_console in src-tauri/src/main.rs). A visible blank console window = regression of that guard (and the user closing it kills the server).