Manage OpenClaw's use of models available at wisgate.ai API

Query Wisgate for model pricing, capabilities, and configuration details. Use when adding a model to OpenClaw, comparing Wisgate pricing, or checking if a mo...

Security Scan

The skill's instructions mostly match its stated purpose (managing Wisgate models), but the SKILL.md references undeclared credentials and hard-coded local paths and expects the agent to read/write local config files — these mismatches should be clarified before trusting it.

Purpose & Capability

The name/description align with the instructions: the skill is about adding and inspecting Wisgate models and mapping them into OpenClaw config. However, the SKILL.md expects access to local OpenClaw files (MODELS.json, openclaw.json, TOOLS.md) and a Firecrawl script in a specific path (~/.openclaw/workspace/skills/...), yet the skill metadata declares no required files, binaries, or environment variables. That mismatch suggests incomplete metadata or sloppy documentation.

Instruction Scope

Instructions tell the agent to read and write local configuration files (MODELS.json, openclaw.json), and to run a local Python scraper script that will fetch pages from https://wisgate.ai. They also reference an environment variable (FIRECRAWL_API_KEY) and TOOLS.md for the key. The skill text does not explicitly limit what other files or env vars may be read; relying on unspecified local files and keys broadens scope beyond what's declared.

Install Mechanism

There is no install spec and no code files (instruction-only), which is the lowest-risk install mechanism. The main risk comes from instructions invoking an external local script rather than from any install process.

Credentials

The SKILL.md uses FIRECRAWL_API_KEY but the skill metadata lists no required environment variables or primary credential. It also expects access to TOOLS.md and to modify MODELS.json and openclaw.json. Requesting these local files/keys would be reasonable for the stated purpose, but the omission from metadata is an incoherence that should be resolved before use.

Persistence & Privilege

The skill is not always-loaded and is user-invocable, which is appropriate. It does instruct modifying agent configuration files (MODELS.json, openclaw.json) — normal for a model-management helper — but you should be aware it expects permission to write those configs.

What to consider before installing

This skill appears to do what it says (manage Wisgate models) but its documentation is incomplete and expects local assets and an API key that are not declared in the metadata. Before installing or letting the agent run it autonomously: 1) Verify the Firecrawl script exists at ~/.openclaw/workspace/skills/firecrawl-search/scripts/scrape.py and inspect its code to ensure it only scrapes the intended pages. 2) Find where FIRECRAWL_API_KEY is stored (TOOLS.md is referenced) and avoid putting unrelated secrets there; prefer declaring required env vars explicitly. 3) Back up MODELS.json and openclaw.json before allowing the skill to edit them. 4) If you do not want config changes made automatically, restrict autonomous invocation or require manual approval for writes. Clarifying these missing metadata items (required env var, config paths, and the Firecrawl dependency) would raise confidence.

Wisgate Model Management

Canonical Model Catalog

Always check MODELS.json first before using Firecrawl or the API:

python3 -c "
import json
with open('MODELS.json') as f:
    data = json.load(f)
# List all models
for k, v in data.items():
    print(k)
"

To look up a specific model:

python3 -c "
import json, pprint
with open('MODELS.json') as f:
    data = json.load(f)
pprint.pprint(data.get('MODEL_ID'))
"

MODELS.json contains what you need for setting up the model in openclaw.json: openclaw_provider_key, api_type, context_window, max_output_tokens, notes (your and Florian's notes on that model), and all pricing fields.

Adding a Model to OpenClaw

Step 1: Look up the model in MODELS.json to get all fields.

Step 2: Add to agents.defaults.models in openclaw.json:

"PROVIDER_KEY/MODEL_ID": {}

e.g. "custom-api-wisgate-ai-openai/qwen3-max": {}

Step 3: Add the full model definition to the provider's models array:

{
  "id": "MODEL_ID",
  "name": "Display Name (Wisgate)",
  "contextWindow": 200000,
  "maxTokens": 33000,
  "input": ["text"],
  "cost": {
    "input": 0.50,
    "output": 2.00,
    "cacheRead": 0.10,
    "cacheWrite": 0.50
  },
  "reasoning": false
}

• reasoning: true only for models with -thinking in the name
• input: ["text"] for LLMs; ["text", "image"] for multimodal

When MODELS.json Is Missing a Model

Use Firecrawl to fetch from Wisgate:

FIRECRAWL_API_KEY=[see_below_for_where_to_find_the_key] \
python3 ~/.openclaw/workspace/skills/firecrawl-search/scripts/scrape.py \
"https://wisgate.ai/models/MODEL_ID"

Extract from output:

• CTX → contextWindow
• Input: **$X.XX** per 1M tokens → cost.input
• Output: **$Y.YY** per 1M tokens → cost.output
• Cache Read: / Cache Write: → cost.cacheRead / cost.cacheWrite

Then also add the new model to MODELS.json so future lookups work.

API Type Mapping

Endpoint Type	OpenClaw Provider	Base URL
Anthropic (Claude)	custom-api-wisgate-ai	https://api.wisgate.ai/
OpenAI compatible	custom-api-wisgate-ai-openai	https://api.wisgate.ai/v1
Google (Gemini)	custom-api-wisgate-google	https://api.wisgate.ai/v1beta

Firecrawl API Key

Stored in TOOLS.md