higgsfield-brief-to-pipelinelisted
Install: claude install-skill lucasyhzhu-debug/higgsfield-brief-to-pipeline
# Higgsfield Brief → Pipeline
## Overview
Turn a **brief + a folder of references** into an **optimal, credit-aware Higgsfield pipeline**,
shown to the user as an **HTML process preview FIRST**, then executed stage-by-stage with
auto-verification, human gates, and zero-credit deterministic overlays for text/dimensions.
Core principle: **spend tokens (planning, verification, overlays) to save credits (renders).**
For the same reason, copy is adversarially reviewed **before** the render, at step 2.5.
## Prerequisites
- The [Higgsfield CLI](https://higgsfield.ai) on `PATH`, authenticated via `higgsfield auth login`.
- Python 3.9+. `build_plan_html.py` is pure stdlib; `overlay.py` needs `pillow` + `numpy`.
## Locating this skill's scripts (RUN THIS FIRST)
The two helper scripts ship **inside this skill**, not in the user's project. `scripts/…` is never
relative to the cwd. **`$CLAUDE_PLUGIN_ROOT` is NOT exported into the Bash tool environment** — it
is only set for hooks and MCP servers — so resolve the directory by probing, once per run:
```bash
SK="${CLAUDE_PLUGIN_ROOT:+$CLAUDE_PLUGIN_ROOT/skills/higgsfield-brief-to-pipeline}"
[ -d "$SK" ] || SK=$(ls -d "$HOME"/.claude/plugins/cache/*/higgsfield-brief-to-pipeline/*/skills/higgsfield-brief-to-pipeline 2>/dev/null | sort -V | tail -1)
[ -d "$SK" ] || SK="$HOME/.claude/skills/higgsfield-brief-to-pipeline"
[ -f "$SK/scripts/build_plan_html.py" ] || { echo "FATAL: cannot locate skill dir (tried: $SK)"; exit 1; }
echo "SK=$SK"