<< All versions

Skill v1.0.0

currentLLM-judged scan90/100
rbrown101010/rileys-skills/quiver-svg
──Details
PublishedOctober 24, 2026 at 08:32 AM
Content Hashsha256:6fcd74536156ef78...
Git SHAb4cc41e4414b
──Files
Files (1 file, 3.7 KB)
SKILL.md3.7 KBactive
SKILL.md · 99 lines · 3.7 KB

version: "1.0.0" name: quiver-svg description: Generate, inspect, and vectorize SVG assets with the QuiverAI API. Use when Codex needs to create SVG logos, icons, illustrations, diagrams, or other vector graphics from text prompts, convert raster images such as PNG/JPEG/WebP into SVG, list Quiver models, or troubleshoot QuiverAI API requests using QUIVERAI_API_KEY.


Quiver SVG

Overview

Use QuiverAI to create production-ready SVGs from text prompts or raster images. Prefer the bundled REST helper when you need a repeatable command-line flow; use the Node SDK only inside projects that already use Node and want Quiver as an application dependency.

Authentication

  • Never write API keys into skill files, repo files, shell history examples, or generated artifacts.
  • Read the key from QUIVERAI_API_KEY first.
  • If QUIVERAI_API_KEY is not set, the helper reads the key from macOS Keychain service codex-quiver-ai.
  • Store a pasted key in Keychain only when the user explicitly asks for durable setup.
  • Treat generated SVG as code before embedding it in an app: inspect it for unexpected scripts, external references, or oversized markup.

Quick Start

If the key is not configured yet, store it in macOS Keychain:

bash
python3 /Users/rileybrown/.codex/skills/quiver-svg/scripts/quiver_svg.py configure-key

Then generate:

bash
python3 /Users/rileybrown/.codex/skills/quiver-svg/scripts/quiver_svg.py generate \
--prompt "Flat monochrome rocket icon, centered, clean geometry" \
--out rocket.svg

For model discovery:

bash
python3 /Users/rileybrown/.codex/skills/quiver-svg/scripts/quiver_svg.py models

For image vectorization:

bash
python3 /Users/rileybrown/.codex/skills/quiver-svg/scripts/quiver_svg.py vectorize \
--image-file input.png \
--auto-crop \
--out input-vectorized.svg

Workflow

  1. Confirm a key is configured with key-status before making API calls when auth is uncertain.
  2. Use models when the user asks what Quiver can use or when model availability may matter.
  3. Use generate for prompt-to-SVG tasks. Default to arrow-1.1; use arrow-1.1-max when the user asks for highest quality or the asset is detail-sensitive.
  4. Use vectorize for PNG, JPEG, or WebP to SVG. Manually crop tightly first when possible; otherwise pass --auto-crop.
  5. Save SVG outputs as explicit files in the working directory or user-requested destination.
  6. Open or inspect the SVG after generation when quality, safety, or dimensions matter.

Prompting

Include the subject, style, palette, and composition. Put production constraints in --instructions.

Good prompt:

text
Minimal vector logo mark for a local courier app: folded paper route arrow forming a C, centered, no text, works at 24px.

Good instructions:

text
Use clean editable SVG structure, no raster images, no external URLs, balanced whitespace, two colors maximum.

Helper Script

The bundled script is at scripts/quiver_svg.py.

Commands:

  • models: calls GET /v1/models.
  • generate: calls POST /v1/svgs/generations.
  • vectorize: calls POST /v1/svgs/vectorizations.

Common options:

  • --model arrow-1.1 or --model arrow-1.1-max
  • --out path.svg for one output or --out output-dir for multiple outputs
  • --raw-response response.json to keep Quiver response metadata
  • --temperature, --top-p, --presence-penalty, and --max-output-tokens when sampling control is needed

For text generation, pass --reference-url for web-hosted reference images or --reference-file for local image references. Arrow 1.1 supports up to 4 references; Arrow 1.1 Max supports up to 16.

API Notes

Read references/quiver-api.md when you need endpoint details, model guidance, error handling, or parameter ranges.

All versions