training-setup

# Training Setup Skill

## Purpose

Automatically validate configuration and guide users through setting up local hotvect algorithm training runs, ensuring data availability, version compatibility, and proper parameters.

**Hotvect Version:** Always use the currently installed hotvect version. Hotvect is backward compatible - never switch versions to match algorithms.

## ⚠️ Configuration Protection Policy

**CRITICAL: Never modify `~/.hotvect/` configuration files.** See `CONFIG_PROTECTION_POLICY.md` for full policy.

This skill reads `~/.hotvect/config.json` for directory paths but never modifies it. Work only in user-specified output directories.

## Configuration

**CRITICAL:** You (Claude) must read `~/.hotvect/config.json` at skill invocation to construct complete `hv train` commands.

**How this works:**
- The `hv train` CLI tool does NOT read config.json
- YOU read config.json and construct the full command line with all arguments
- Config provides defaults that can be overridden by user-specified values

**When this skill activates:**
1. Read config: `cat ~/.hotvect/config.json`
2. Parse the JSON to extract values (you can understand JSON directly)
3. Use config values as defaults when user hasn't specified:
   - `directories.data_base_dir` → `--data-base-dir` argument
   - `directories.output_base_dir` → `--output-base-dir` argument
   - `aws.credential_helper` → Command for AWS credential refresh

4. **Fail fast** if config doesn't exist and user didn't provide required paths
5. Tell user to run `/agent hotvect-setup-agent` if config is missing
6. **Activate virtual environment**: `source ${hotvect_source_dir}/python/.venv/bin/activate`

**Important:** Always construct complete `hv train` commands with ALL required arguments explicitly specified.

## When to Invoke

Invoke this skill when:
- User mentions training an algorithm
- User is setting up a training run
- User wants to run `hv train`
- User asks about training configuration

## Context Requiremen