FluxLoop Setup Skill

Setup-First: Check state → Install → Authenticate → Select project → Guide next step

Output Format

📎 All user-facing output must follow: read skills/_shared/OUTPUT_FORMAT.md

Context Protocol

setup is a pre-test-memory step — no reads, no writes.

1. fluxloop context show → check current state (determine if setup is already complete)
2. If already complete → guide to context or scenario skill

📎 Full protocol: read skills/_shared/CONTEXT_PROTOCOL.md

Prerequisite

None — this is the entry point. No prior skills required.

Workflow

Step 1: CLI Installation Check

Check if installed first — this must run before any other `fluxloop` command:

fluxloop --version

If not installed (command not found), guide based on the workspace environment:

Priority: Install in the SAME environment where the agent runs — this is critical for simulation.

#### Dependency Safety Guardrails

1. Install fluxloop-cli only (NOT fluxloop)
2. Before install/upgrade, run python --version and uv pip index versions fluxloop-cli
3. If the target version is not published, report available versions and stop (never guess ranges)
4. If Python requirement is higher than the current interpreter, stop and offer two choices: upgrade Python or pin a compatible version
5. On failure, explain root cause in one line (name/version/python mismatch)

Step 2: Context Check

After CLI is confirmed installed, determine current state:

fluxloop context show

Also run fluxloop auth status to verify login state.

Step 3: Authentication

fluxloop auth login --no-wait && fluxloop auth login --resume

These two commands form a single flow:

• --no-wait prints the device code for the user to enter
• --resume polls until the login completes

📎 Staging environment: read skills/_shared/STAGING.md

Step 4: Project Selection

List existing projects:

fluxloop projects list --format json

If a project exists: select it:

fluxloop projects select <project_id>

If creating a new project:

#### Naming Rules

Suggest 3 candidates based on the codebase:

• Analyze: package.json, pyproject.toml, main agent file, README, directory name
• Folder: keywords → kebab-case → suffix (-agent, -bot)
• Display: concise, user-friendly names

📁 Folder: 1) order-bot  2) food-agent  3) cs-service
📝 Name: 1) Order Bot  2) Support Bot  3) Food Agent
Select number or type custom:

#### Language Selection

Ask the user for the default language once:

🌐 Language: 1) en (English)  2) ko (Korean)  3) ja (Japanese)  4) other
Select number or type language code (default: en):

Create the project:

fluxloop projects create --name "my-agent" --language <code>

Required link output: After project creation/selection, extract project_id from CLI output and display: ✅ Project → "{name}" (proj_xxx) 🔗 https://alpha.app.fluxloop.ai/simulate/scenarios?project={project_id}

Error Handling

Next Steps

Setup complete. Available next actions:

• Codebase scan & agent profile creation (context skill)
• Scenario creation (scenario skill — auto-runs context if no profile exists)

Quick Reference

📎 Full CLI reference: read skills/_shared/QUICK_REFERENCE.md

Key Rules

1. Always check fluxloop --version first — if not installed, start from Step 1 (installation)
2. After CLI confirmed, run fluxloop context show — skip to the right step based on state
3. Install fluxloop-cli (NOT fluxloop)
4. Install in the SAME environment where your agent runs
5. Never guess package versions — check uv pip index versions fluxloop-cli
6. Folder names: English kebab-case only (order-bot)
7. Project/scenario display names: any language allowed
8. Suggest 3 naming candidates, allow custom input
9. Ask language once at project creation (default: en)
10. After setup, guide to context skill — never jump to scenario directly