install-app

# Install App Install one application using a structured, verifiable workflow contract. This skill is a generic execution wrapper for app-specific install instructions. Use it with per-app facts from `apps/<app>/` and official upstream docs. Supporting files in this skill: - `checklist.md` - `report-template.md` ## Inputs - app name (required) - install contract/prompt (required): concrete command sequence and success criteria - install target (required before execution): `local` or `remote` - execution mode (optional): `plan_only` | `execute_with_confirmation` | `execute_directly` (default: `execute_with_confirmation`) - language (optional, default `en`) - workspace folder chosen by user (optional at planning, required before execution when commands write files) ## Steps 1. Parse the install contract into explicit phases: prerequisites, install, conflict handling, validation, and user handoff. 2. Ask whether to install on `local` or `remote` host before execution. If the user does not specify, ask explicitly instead of assuming. 3. Run prerequisite checks exactly as required by the contract (for example, `docker compose version`). If a mandatory dependency is missing, stop and report the exact blocker. 4. Resolve required server ports before installation: probe occupancy first, then choose available ports with this priority order: - contract-required fixed port (if mandatory and free) - nearest available port in `9000-9999` - contract fallback strategy (for example `--port <free-port>`) 5. In `execute_with_confirmation` mode, present the exact command plan (including selected target and ports) and wait for user confirmation before executing. 6. Execute install commands in the user-selected folder. Keep commands deterministic and non-interactive when possible. 7. Run all required health checks and capture evidence (for example, `docker compose -p <project> ps` and HTTP health endpoint checks). 8. Produce a concise outcome report using `report-template.md`. 9. Provide handoff details: final access URL, credentials retrieval path (if applicable), and backup reminders (such as backing up generated `.env`). ## Output - install result (`passed` / `failed` / `blocked`) - install target (`local` / `remote`) - execution folder - commands run - port allocation result (required port, occupancy check, and final port) - health check evidence - access URL and next actions ## Rules - Never invent prerequisites, flags, ports, or health endpoints that are not in the contract or repository facts. - Always ask for install target (`local` or `remote`) before execution when it is not explicitly given. - For server-facing ports, prefer available ports in the `9000-9999` range whenever the contract allows custom ports. - Always check port occupancy before install commands that bind ports. - If critical input is missing (for example, target folder or required secret), stop and ask the user before execution. - Do not silently ignore a failed check; surface the first blocking error and its command output. - Keep changes scoped to the target app/work folder. - Match output language to user preference.