QuiverAI

Generate and vectorize SVG graphics via the QuiverAI API (Arrow model). Use when the user asks to create logos, icons, or illustrations as SVG, convert raste...

Security Scan

The skill's requests and runtime instructions match its description: it is an instruction-only wrapper for the QuiverAI SVG/vectorization API and only requires a single API key.

Purpose & Capability

Name/description, endpoints, and example SDK usage all describe generating and vectorizing SVGs via QuiverAI. The single required environment variable (QUIVERAI_API_KEY) is exactly the credential needed to call the documented API.

Instruction Scope

SKILL.md only contains curl/SDK examples and parameter/response docs for api.quiver.ai; it does not instruct reading unrelated files, accessing other environment variables, or sending data to third-party endpoints beyond the QuiverAI API.

Install Mechanism

There is no install spec and no code files (instruction-only). The doc mentions an optional public npm SDK (@quiverai/sdk) but does not download or execute arbitrary code — this is low-risk and proportionate.

Credentials

Only QUIVERAI_API_KEY is required and declared as the primary credential. That single key is appropriate for an API-integration skill; there are no unrelated secrets or excessive env requirements.

Persistence & Privilege

always is false, no config paths or system modifications are requested, and the skill does not ask to persist beyond normal operation. disable-model-invocation is default (false) which is expected and not flagged alone.

Assessment

This skill appears internally consistent and low-risk because it only documents calls to api.quiver.ai and requires a single QuiverAI API key. Before installing: 1) Confirm you trust Quiver.ai and understand their billing (requests may consume credits). 2) Treat QUIVERAI_API_KEY like any secret — provide only to agents you trust, rotate it if shared, and avoid embedding it in public places. 3) Be aware that any raster images you send will be uploaded to the QuiverAI service (check their privacy/terms if images are sensitive). 4) If you prefer, create a limited/throwaway API key for testing. The scanner had no code to analyze (instruction-only), so disk execution risk is low — but runtime network calls to the QuiverAI API will occur when the skill is used.

QuiverAI — AI Vector Graphics

QuiverAI generates production-ready SVGs from text prompts or raster images.

• Site: https://quiver.ai
• Docs: https://docs.quiver.ai
• API base: https://api.quiver.ai/v1
• Model: arrow-preview
• Auth: Bearer token via QUIVERAI_API_KEY
• Billing: 1 credit per request (regardless of n).

Setup

Get an API key at https://app.quiver.ai/settings/api-keys (create account at https://quiver.ai/start first).

Text to SVG

Generate SVGs from a text description.

Endpoint: POST /v1/svgs/generations

curl -X POST https://api.quiver.ai/v1/svgs/generations \
  -H "Authorization: Bearer $QUIVERAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "arrow-preview",
    "prompt": "A minimalist monogram logo using the letter Q",
    "n": 1,
    "stream": false
  }'

Node.js SDK (npm install @quiverai/sdk):

import { QuiverAI } from "@quiverai/sdk";
const client = new QuiverAI({ bearerAuth: process.env.QUIVERAI_API_KEY });

const result = await client.createSVGs.generateSVG({
  model: "arrow-preview",
  prompt: "A minimalist monogram logo using the letter Q",
});
// result.data[0].svg contains the SVG markup

Parameters

Param	Type	Default	Description
model	string	—	Required. Use arrow-preview.
prompt	string	—	Required. Describes the desired SVG.
instructions	string	—	Additional style guidance (e.g. "flat monochrome, rounded corners").
references	array	—	Up to 4 reference images ({ url } or { base64 }).
n	int	1	Number of outputs (1–16).
temperature	float	1	Sampling temperature (0–2). Lower = more deterministic.
top_p	float	1	Nucleus sampling (0–1).
max_output_tokens	int	—	Upper bound for output tokens (max 131072).
stream	bool	false	SSE streaming (events: reasoning, draft, content).

Response

{
  "id": "resp_01J...",
  "created": 1704067200,
  "data": [{ "svg": "<svg ...>...</svg>", "mime_type": "image/svg+xml" }],
  "usage": { "total_tokens": 1640, "input_tokens": 1200, "output_tokens": 440 }
}

Image to SVG (Vectorize)

Convert a raster image (PNG/JPEG/WebP) into SVG.

Endpoint: POST /v1/svgs/vectorizations

curl -X POST https://api.quiver.ai/v1/svgs/vectorizations \
  -H "Authorization: Bearer $QUIVERAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "arrow-preview",
    "stream": false,
    "image": { "url": "https://example.com/logo.png" }
  }'

SDK:

const result = await client.vectorizeSVG.vectorizeSVG({
  model: "arrow-preview",
  image: { url: "https://example.com/logo.png" },
});

Additional parameters (beyond Text-to-SVG shared ones)

Param	Type	Default	Description
image	object	—	Required. { url: "..." } or { base64: "..." }.
auto_crop	bool	false	Crop to dominant subject before vectorization.
target_size	int	—	Square resize target in px (128–4096) before inference.

Response format is identical to Text-to-SVG.

Error codes

Status	Code	Meaning
400	invalid_request	Malformed body or missing fields.
401	unauthorized	Bad or missing API key.
402	insufficient_credits	Out of credits.
429	rate_limit_exceeded	Too many requests; back off and retry.

Tips

• Save SVG output to a .svg file for immediate use.
• Use instructions to control style without changing the prompt.
• For logos, try low temperature (0.3–0.5) for cleaner, more consistent results.
• Use references to provide visual examples the model should match.
• For vectorization, enable auto_crop: true when the source image has excess whitespace.