setup

TinyML Agent Skill — Setup

Run this once after installing the plugin. Re-run any time you move or reinstall tinyml-modelzoo, or want to change your update mode.

---

Step 1: Discover script paths

Two separate runner.py files exist — one for this setup skill, one for the main tinyml-workflow-agent skill. Find both:

Main skill scripts (SCRIPTS_DIR) — used for tinyml-modelzoo operations during this session only (not stored in .env):

find ~/.claude -name "runner.py" 2>/dev/null | grep "tinyml-workflow-agent" | head -1

Set SCRIPTS_DIR from result (strip /runner.py, keep the directory).

Setup skill scripts (SETUP_SCRIPTS_DIR) — used only during this setup:

find ~/.claude -name "runner.py" 2>/dev/null | grep "setup/scripts" | head -1

Set SETUP_SCRIPTS_DIR from result (strip /runner.py, keep the directory).

If either is not found, ask the user:

"Where is the tinyml-agent-skills plugin installed?"

Verify both runners exist:

ls "$SCRIPTS_DIR/runner.py"
ls "$SETUP_SCRIPTS_DIR/runner.py"

---

Step 2: Choose update mode

Ask the user:

"How would you like to manage updates for this skill?

1. Pinned — stay on the current version, no automatic updates
2. Auto-update — check for newer versions at the start of each session"

NOTE: Current version can be found in plugins/tinyml-agent-skills/.claude-plugin/plugin.json,

Call the setup runner (not the main skill runner) with their choice:

# Pinned:
UPDATE_RESPONSE=$(python3 "$SETUP_SCRIPTS_DIR/runner.py" set_update_mode '{"mode": "pinned"}')

# Auto-update:
UPDATE_RESPONSE=$(python3 "$SETUP_SCRIPTS_DIR/runner.py" set_update_mode '{"mode": "auto"}')

Confirm success: true from UPDATE_RESPONSE before proceeding.

Extract and store for Step 7:

UPDATE_MODE=$(echo "$UPDATE_RESPONSE" | python3 -c "import sys, json; print(json.load(sys.stdin).get('mode'))")
UPDATE_PINNED_VERSION=$(echo "$UPDATE_RESPONSE" | python3 -c "import sys, json; print(json.load(sys.stdin).get('pinned_version') or '')")