Yieldingbear

Use Yielding Bear's unified LLM API for cost arbitrage and intelligent routing. Use when cutting AI costs, routing LLM requests, comparing model pricing, or...

Security Scan

The skill's behavior (routing prompts to a third‑party API) matches its description, but the runtime instructions ask you to store and use an API key while the registry metadata does not declare any required credential; the source/homepage are unknown and the README suggests persisting secrets in shell profiles — these inconsistencies and privacy implications warrant caution.

Purpose & Capability

The skill claims to be a unified LLM routing API and the SKILL.md shows exactly the network calls and usage you'd expect for that purpose (curl examples, an OpenAI-compatible base_url). However the registry metadata lists no required environment variables or primary credential while the instructions clearly require YIELDINGBEAR_API_KEY — this mismatch is unexpected and reduces trust. The source and homepage are marked unknown/none which prevents easy verification of the vendor.

Instruction Scope

Runtime instructions tell the agent/user to set and optionally append an API key to ~/.zshrc, and state the agent 'uses YB automatically when calling OpenAI-compatible endpoints.' That grants the third party broad access to any prompts or data sent through OpenAI-compatible flows. Persisting API keys in shell profiles is insecure and the automatic routing claim is vague — it could cause unexpected exfiltration of sensitive prompts if enabled without clear controls.

Install Mechanism

This is an instruction-only skill with no install spec and no code files, so there is no installer or downloaded binary to review. That minimizes filesystem/installation risk, but means all risk is from network calls and the runtime instructions.

Credentials

The SKILL.md requires a single API key (YIELDINGBEAR_API_KEY) — reasonable for a third‑party API — but the registry does not declare this required env var or primary credential. That inconsistency is suspicious because users won't be warned at install time that a secret is needed. Additionally, instructions recommend storing the key in shell startup files (persistence of secrets in plaintext), which is disproportionate from a security best-practices perspective.

Persistence & Privilege

The skill does not request 'always: true' and does not include install-time changes or system-wide configuration. It simply depends on an API key at runtime. Autonomous invocation is allowed by default but not a unique escalation here.

What to consider before installing

This skill routes your prompts through a third-party service and requires an API key according to its instructions — but the registry metadata does not declare that key, and the publisher/homepage are unknown. Before using it: 1) Verify yieldingbear.com and the vendor (look for a privacy policy, docs, and reputation). 2) Never paste sensitive data or production prompts while testing — try harmless inputs first. 3) Avoid persisting secrets in shell rc files; prefer ephemeral env vars or a secrets manager. 4) If you plan to let agents call OpenAI-compatible endpoints, confirm how and when YB will be invoked so you don't accidentally route sensitive requests. 5) Ask the publisher or registry maintainer to correct the missing required env var (YIELDINGBEAR_API_KEY) in metadata and to provide a verified source/homepage. If you cannot validate the vendor and its policies, treat this skill as risky and avoid exposing confidential data to it.

Yielding Bear — Unified LLM Routing API

Yielding Bear provides a single unified API that routes every LLM request to the cheapest capable model across 16+ providers — saving 60-80% vs calling OpenAI, Anthropic, or Google directly.

Setup (First Time Only)

1. Get an API key at https://yieldingbear.com/api

2. Set environment variable:

   export YIELDINGBEAR_API_KEY="yb_live_your_key_here"
   

3. Save to your shell profile (optional):

   echo 'export YIELDINGBEAR_API_KEY="yb_live_your_key_here"' >> ~/.zshrc
   source ~/.zshrc
   

Quick Start — OpenClaw Agents

Method 1: Direct API calls

curl -X POST https://api.yieldingbear.com/v1/chat/completions \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [{"role": "user", "content": "Summarize this email: ..."}],
    "max_tokens": 500
  }'

Method 2: OpenAI-compatible drop-in replacement

from openai import OpenAI

client = OpenAI(
    api_key="yb_live_your_key",
    base_url="https://api.yieldingbear.com/v1"
)
# Same SDK. Same code. 60-80% less cost.

Model Routing

Task Type	Routes To	Cost/1M
Summaries, classification	Llama 3.1 8B	$0.04
Email drafting, formatting	DeepSeek V3	$0.07
General chat, code	GPT-4o-mini	$0.15
Complex reasoning	YB Sentinel 70B	$0.06
Fast completions	Gemini 2.0 Flash	$0.10

Override routing:

{ "model": "claude-3.5-haiku", "routing": { "capabilities": ["reasoning"] } }

OpenClaw Agent Integration

For OpenClaw sub-agents

Set in environment before spawning:

export YIELDINGBEAR_API_KEY="yb_live_..."

The agent uses YB automatically when calling OpenAI-compatible endpoints.

For custom tools and scripts

RESULT=$(curl -s -X POST "https://api.yieldingbear.com/v1/chat/completions" \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"auto","messages":[{"role":"user","content":"Analyze: $1"}]}')
echo "$RESULT"

Cost Comparison

Task	Direct OpenAI	Via YB	Savings
1M simple summaries	$150	$40	73%
1M email drafts	$300	$60	80%
1M chat completions	$500	$150	70%
1M reasoning tasks	$3,000	$300	90%

Key Links

• API Docs: https://yieldingbear.com/developers
• Yields page: https://yieldingbear.com/yields
• How it works: https://yieldingbear.com/how-it-works
• Dashboard: https://yieldingbear.com/dashboard
• Get API key: https://yieldingbear.com/api