creator-breakout-kit: Instructions file for Claude Code

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

What this repo is

Creator Breakout Kit is the product that finds the creator angles, hooks, and concepts most likely to break out for a brand — before the brand pays to source or produce. Strategy before sourcing, pattern fit over follower count. The final shareable artifact is the Creator Breakout Brief (creator-breakout-brief.md and creator-breakout-brief.html).

Internally, it's a Python + prompt package: no app, server, or build step — just a pipeline of deterministic scripts, thin connector adapters, and skill prompt modules stitched together via a shared run directory.

One brand = one runs/<date>-<slug>/ directory. Every stage reads and writes files in that directory; do not introduce global state or cross-run coupling.

Common commands

# Setup / check connector env
cp .env.example .env
bash scripts/preflight.sh

# Tests (stdlib unittest, no pytest, no package __init__.py files).
# All test commands must be run from the repo root.
python3 -m unittest discover -s tests -v                                # all tests
python3 tests/test_enrich_breakout_run.py -v                            # one file
python3 tests/test_enrich_breakout_run.py TestEnrichBreakoutRun.test_one_sentence_preserves_abbreviations -v  # one method

There is no linter, formatter, or package manager — all scripts use only Python 3 stdlib plus optional HTTP calls inside adapters. Don't add a requirements.txt, pyproject.toml, or framework dependency without discussing first.

Pipeline architecture

The end-to-end flow moves through three layers that share a run directory:

1. Base run — scripts/create_base_breakout_run.py writes input.json, breakout-kit.json, creator-breakout-brief.md (the Creator Breakout Brief), internal_notes.md, (optional) sales_handoff.md. This alone is the quick-pass mode and requires no connectors.
2. Connector research — scripts/run_creator_breakout_research.sh drives skills/brand-snapshot/scripts/firecrawl_brand_discovery.py and skills/breakout-pattern-research/scripts/virlo_outlier_research.py into a separate $RESEARCH_DIR, then scripts/enrich_breakout_run.py folds those outputs back into the run directory. Optional stages: skills/breakout-pattern-research/scripts/virlo_tracking_intelligence.py (audience-signal pass) and skills/breakout-pattern-research/scripts/video_intelligence.py (ScrapeCreators + OpenRouter Gemini video asset review). Adapter scripts live inside the skill that uses them, per the pattern in the other kits (landing-page-factory, meta-ads-kit) — each skill is self-contained with its own scripts/ and references/ subdirectories.
3. Synthesis — scripts/synthesize_full_kit.py is the only stage allowed to produce the shareable final brief. It materializes brand-snapshot.json, breakout-pattern-research.json, creator-archetype-map.json, concept-slate.json, skill-run-manifest.json, refreshes the brief/notes/handoff (markdown), and also emits creator-breakout-brief.html — a dark-terminal-aesthetic presentation-layer version of the same brief rendered by scripts/render_breakout_brief_html.py. Both the markdown and HTML renderers read from the same breakout-kit.json and share concept_scores / one_sentence via importlib, so score labels and first-test copy never drift between formats. Synthesis imports both enrich_breakout_run and render_breakout_brief_html dynamically — keep both modules importable as scripts (no top-level side effects at import time).