Skip to main content

openspec-dashboard

Open, set up, or inspect the local OpenSpecUI dashboard for an OpenSpec project. Check required CLI dependencies first, install or initialize missing pieces when authorized, and keep the dashboard read-only unless the user explicitly requests an action.

Informations de source

Dépôt
SamuelAsherRivello/ai-skills-library
Dernière activité de la source
12 octobre 2026 à 09:11
Langue détectée de SKILL.md
anglais
Étoiles
0
Forks
0

Options d'installation

Le prompt qui vérifie d'abord la source est sélectionné par défaut. Vous pouvez passer à une commande directe ou télécharger une copie locale.

Vérifiez les fichiers source

Lisez SKILL.md et les fichiers associés affichés par SkillsMP avant de décider de l'installer.

Explorateur de fichiers
2 fichiers

Affichage de SKILL.md

SKILL.md
Instructions source · Aperçu en lecture seule
name
openspec-dashboard
description
Open, set up, or inspect the local OpenSpecUI dashboard for an OpenSpec project. Check required CLI dependencies first, install or initialize missing pieces when authorized, and keep the dashboard read-only unless the user explicitly requests an action.
# OpenSpec Dashboard ## Chat title Follow the shared [chat title contract](../openspec-title-management.md). This workflow uses `OS-dsh: {change}` and renders the shared footer. Use this skill when the user asks to open, inspect, repair, or use the local OpenSpecUI dashboard for an OpenSpec project. This skill is a local dashboard launcher and inspector. It does not call an AI service, delegate work to another agent, or invent a separate dashboard. When invoked for a dashboard request, use the terminal before replying; do not only print commands for the user to run. When the request also concerns the project's Vite preview server, use the project lifecycle skills as needed: - `ai-skills-project-run-start` for starting or restarting the project's Vite server and verifying its application routes. - `ai-skills-project-run-status` for read-only project-server/process/route status. - `ai-skills-project-run-stop` for stopping only the current project's Vite server. Resolve these skills from the current project's `.agents/skills/` first, then from the installed global skills. Do not use the Vite lifecycle skills to start, stop, or inspect the OpenSpecUI server itself. Keep the dashboard and project preview as separate servers, and preserve each skill's ownership and one-server invariants. ## Dependency And Project Setup Before launching the dashboard, check the current project and required CLIs. 1. Resolve the project path from the user's request or the current working directory. 2. Resolve the OpenSpec root with `openspec context --json` (or `openspec list --json` when that is the supported compatibility path). Use the returned `root.path` as canonical rather than inferring a `.openspec/` or `openspec/` folder. A null root means initialization would be a planning write: explain that and wait for explicit approval. If a declared store cannot be resolved, stop and report the CLI's concrete error instead of treating the project as uninitialized. 3. Check CLI availability with read-only commands: - `openspec --version` - `openspecui --help` 4. If OpenSpec is missing, inspect the project documentation for its expected install command. If no project-specific command exists, ask for approval before installing the OpenSpec CLI globally with npm. 5. If OpenSpecUI is missing, ask for approval before installing the dashboard CLI globally with npm. 6. If the project is not initialized for OpenSpec, explain that initialization will create or update project planning files and ask for explicit approval before running `openspec init` or an equivalent setup command. 7. If setup or installation fails, report the exact failing command and concrete error. Do not claim the dashboard is unavailable without first attempting the relevant check or setup path. Prefer existing local commands and project scripts over global installs. Never ask the user to paste secrets. Do not modify implementation code as part of dashboard setup. ## Launch Use browser-safe port `6001` by default; Chromium blocks port `6000` as unsafe. For a project path: ```powershell openspecui serve "<project-path>" --web --port 6001 --no-open ``` If port `6001` is already in use, choose the next available safe localhost port and report the actual URL. The dashboard URL is usually: ```text http://localhost:6001 ``` Leave the server running for the preview session. If browser control is available, open or focus the dashboard and inspect the visible page before reporting completion. If the user asks to run the project preview alongside the dashboard, invoke `ai-skills-project-run-start` for the Vite server after the dashboard is available, then report both URLs and clearly label which server each URL uses. Do not open project routes merely because the dashboard was launched. If the user asks whether the project is running while using the dashboard, invoke `ai-skills-project-run-status` and report its read-only findings. If the user asks to stop the project preview, invoke `ai-skills-project-run-stop`. Stopping the project preview must not stop OpenSpecUI unless the user explicitly asks to stop the dashboard too. For the native application surface, use only when the user asks for it: ```powershell openspecui --app ``` Use `openspecui --help` when the user asks about available modes or when the launch command fails. ## Read-Only Inspection Treat the dashboard as a view by default. When the user asks to look at the dashboard, summarize the actual visible state: active changes, artifact/task progress, blocked or ready work, and repository scope. Do not summarize the product marketing page when the project dashboard is what the user requested. Do not click workflow actions such as Apply, Archive, Continue, Fast-forward, Verify, or destructive/irreversible controls unless the user explicitly asks for that action. ## Configure Codex As The Default Agent When the user explicitly asks to configure the dashboard or project for Codex, use the dashboard's **Settings → Agent Integrations → Manage** surface. Select **Codex** in Agent Inventory and run the server-owned initialization command. OpenSpecUI currently invokes: ```powershell openspec init --tools codex --profile custom ``` This creates or refreshes the project-local `.agents/skills/` OpenSpec skills and records `codex` in `.agents/skills/.openspec-target`; it does not install the Codex application or modify product implementation code. Treat this as a project-planning configuration write: do it only when the user explicitly asks for Codex setup, then verify the command exits successfully and the target file contains `codex`. Reload the dashboard before reporting the result. ## Windows CLI-Runner Recovery If the dashboard reports `where.exe EPERM`, `spawn EPERM`, or cannot locate the OpenSpec runner: 1. Verify `where.exe openspec`, `where.exe openspecui`, and `node --version`. 2. Find the installed OpenSpec JavaScript runner without printing secrets. Common npm global locations are under `%APPDATA%\npm\node_modules\@fission-ai\openspec\bin\openspec.js`. 3. If the dashboard has an Execute Path setting, set it to the direct Node invocation for the discovered runner, for example: ```text "C:\Program Files\nodejs\node.exe" "C:\Users\<user>\AppData\Roaming\npm\node_modules\@fission-ai\openspec\bin\openspec.js" ``` 4. Restart the dashboard from a terminal with permission to spawn child processes. 5. Reload the dashboard and confirm the CLI error is gone before reporting success. If elevated execution is required, request approval for the specific command instead of silently changing permissions or substituting another dashboard.
Voir sur GitHub