Installation

Clone the repo. Open Claude Code. Start asking questions.

Prerequisites

• Claude Code -- the CLI that runs the orchestrator. Install with npm install -g @anthropic-ai/claude-code (requires Node.js 18+). See the Claude Code docs.
• Python 3.10+ -- all scripts are stdlib only. No third-party packages needed at runtime.
• macOS or Linux -- POSIX shell. Windows via WSL.

Install

git clone https://github.com/kamilseghrouchni/vcro-sourcing.git && cd vcro-sourcing

That's it. The repo contains everything: the CLI, agents, skills, rules, hooks, scripts, and a pre-built wiki with 335 entities. No pip install. No build step.

Verify

# Check version
python3 bin/vcro version

# Browse the wiki
python3 bin/vcro wiki browse

# Validate all entities against the schema
python3 bin/vcro wiki verify

Start using it

# Open Claude Code in the repo directory
claude

# Ask a question -- the orchestrator handles everything
Find AD CSF DNA methylation cohorts with n>100 case-control

# Or use a slash command
/source IBD stool shotgun metagenomics cohorts, treatment-naive, n>=100

Claude Code reads CLAUDE.md and the agent definitions in .claude/agents/ on startup. The orchestrator (vcro-os) routes your question to the right workflow, spawns Sonnet subagents for the heavy work, and delivers a scored recommendation.

What gets used

vcro-sourcing/
  bin/vcro                  # CLI entry point (optional TUI with rich/prompt_toolkit)
  scripts/                  # Pure Python plumbing (ingest, index, lint, telemetry)
  .claude/
    agents/                 # vcro-os, vcro-bounty, vcro-onboard (Opus orchestrators)
    skills/                 # 17 phase-level skills (Sonnet workers)
    rules/                  # 9 rule files + commandments
    hooks/                  # Entity schema validation, reindex, commit authorship
    commands/               # /source, /bounty, /onboard, /lint, /compile
  store/
    wiki/                   # 335 pre-built entity articles
    raw/                    # Immutable source documents
    queries/                # Query run artifacts (audit trail)
    runs/                   # Compile telemetry
  CLAUDE.md                 # Project instructions for Claude Code

Advanced: install.sh

For deploying viCRO into an existing Claude Code project without cloning the full repo, use the installer script:

# From a local checkout
bash scripts/install.sh --from-local /path/to/vcro-sourcing --target ./my-project

# Skills only (no wiki skeleton)
bash scripts/install.sh --from-local /path/to/vcro-sourcing --skills-only --target ./my-project

See bash scripts/install.sh --help for all flags.