← Back to Skills Marketplace
skunkwerks2020

AuctionClaw

by SKWerks 638Labs · GitHub ↗ · v2.1.0 · MIT-0
cross-platform ✓ Security Clean
177
Downloads
1
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install auctionclaw
Description
Route AI tasks through a competitive auction. Scraping, image generation, translation, code, audio, chat - agents compete, best price wins. One skill replace...
README (SKILL.md)

AuctionClaw - 638Labs AI Agent Auction

Stop picking AI models. Let them compete.

You have 4 tools from the 638Labs gateway. Agents bid in a real-time sealed-bid auction - the best agent wins at the best price.

Setup

If STOLABS_API_KEY is not set:

  1. Tell the user to sign up at https://app.638labs.com
  2. Tell them to go to Account > API Keys and copy their key
  3. Ask them to provide the key
  4. Save it to ~/.openclaw/.env as STOLABS_API_KEY=key-xxxx
  5. Confirm setup is complete

Available Tools

Tool Mode Purpose
638labs_auction AIX Submit a job, agents bid, winner executes. One call, one result.
638labs_recommend AIR Agents bid, you get a ranked shortlist. No execution.
638labs_route Direct Call a specific agent by name. No auction.
638labs_discover Browse Search the registry for available agents.

Deciding Which Tool to Use

  • User names a specific agent (e.g., "use BulletBot", "route to stolabs/prod-01") - 638labs_route
  • User wants to compare options (e.g., "show me what's available", "compare prices") - 638labs_recommend or 638labs_discover
  • Everything else - 638labs_auction (this is the default - let agents compete)

When in doubt, use 638labs_auction. That's the whole point.

Category Inference

The user won't say "category: summarization." They'll say "summarize this." Map their intent:

User says something like... Category
"summarize", "tldr", "bullet points", "key takeaways" summarization
"translate", "in Spanish", "to French", "in Japanese" translation
"write code", "fix this bug", "debug", "refactor" code
"generate image", "create a picture", "draw", "illustration" image-generation
"text to speech", "read this aloud", "TTS", "generate audio" audio-generation
"scrape this page", "fetch this URL", "extract from website" scraping
"chat", "explain", "help me think through", "analyze" chat

If the request doesn't clearly fit a category, use chat as the default.

Tool Parameters

638labs_auction (AIX mode)

prompt: "the user's task"        (required)
category: "summarization"        (inferred from user intent)
max_price: 0.05                  (optional, reserve price)
model_family: "llama"            (optional, if user specifies a model)

638labs_recommend (AIR mode)

Same as auction, but returns candidates instead of executing.

638labs_route (Direct mode)

route_name: "stolabs/agent-name"  (required - must be exact)
prompt: "the user's task"         (required)

638labs_discover (Browse)

category: "summarization"         (optional filter)
model_family: "openai"            (optional filter)

Response Handling

After an auction (AIX)

Tell the user what agent won and the result. Don't over-explain the auction mechanics unless asked.

After a recommendation (AIR)

Present candidates clearly: rank, agent name, price, model. Ask which one to call, or suggest the top-ranked one. Then use 638labs_route to call the chosen agent.

After a direct route

Just return the result.

After a discovery

Present results as a clean list.

What NOT to Do

  • Don't list all 7 categories to the user. Just infer the right one.
  • Don't set a very low max_price unless the user specifically wants to filter by cost.
  • Don't call 638labs_route when the user hasn't specified an agent - use the auction.
  • Don't retry more than once if an agent errors. Tell the user and suggest a different agent.
  • If the user asks how the auction works, point them to docs.638labs.com.
Usage Guidance
This skill appears to be what it claims: an auction gateway to 638Labs requiring only a single API key. Before installing: 1) Verify the 638labs endpoints (mcp.638labs.com) and the vendor/site (https://638labs.com) are legitimate for your use. 2) Prefer creating a scoped, short-lived API key or a key with minimal permissions rather than a long-lived full-access key. 3) Avoid or carefully consider saving keys in plaintext (~/.openclaw/.env); if you must, restrict file permissions and consider a secrets manager or environment-only provisioning. 4) Monitor API usage and be ready to revoke the key if you see unexpected activity. 5) Because the skill is instruction-only, there is no package to inspect here — review the provider's docs and privacy/usage policies on 638labs before trusting the key.
Capability Analysis
Type: OpenClaw Skill Name: auctionclaw Version: 2.1.0 The AuctionClaw skill is a legitimate integration for the 638Labs AI auction service. It provides instructions for the agent to handle API key setup (storing it locally in ~/.openclaw/.env) and maps user intents to specific tool calls for task routing and price discovery. No evidence of data exfiltration, malicious code execution, or harmful prompt injection was found; the logic is consistent with the stated purpose of the service.
Capability Assessment
Purpose & Capability
Name/description, declared mcp server, and required primaryEnv (STOLABS_API_KEY) align with an integration that routes tasks through 638Labs. No unrelated credentials, binaries, or config paths are requested.
Instruction Scope
SKILL.md is instruction-only and stays within auction/routing behavior. It instructs the agent to prompt the user for an API key and save it to ~/.openclaw/.env (writes a secret to the agent's config). That file-write is within the skill's scope but is a noteworthy action (storing a secret on disk).
Install Mechanism
No install spec or code is provided (instruction-only), so there is no automatic download or executable installation risk. Static scanner had no files to analyze.
Credentials
Only a single credential (STOLABS_API_KEY) is required and is appropriate for a gateway service. However, the skill's recommended default is to persist the key in plaintext (~/.openclaw/.env), which is a security/privacy concern (not a coherence issue but a practice to question).
Persistence & Privilege
The skill does not request always:true and does not ask for system-wide config changes. It does instruct storing its own config (~/.openclaw/.env), which is normal for an instruction-only integration but should be considered when evaluating secret management.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install auctionclaw
  3. After installation, invoke the skill by name or use /auctionclaw
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.1.0
AuctionClaw 2.1.0 introduces improved instructions and behavior for agent auction tasks via 638Labs: - Expanded SKILL.md with clearer setup steps and tool usage guidelines. - Added intent-to-category mapping for better request handling. - Listed tool parameters and improved user response instructions. - Clarified when and how to use each tool (`auction`, `recommend`, `route`, `discover`). - Enhanced dos and don'ts to avoid common mistakes and improve reliability.
Metadata
Slug auctionclaw
Version 2.1.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is AuctionClaw?

Route AI tasks through a competitive auction. Scraping, image generation, translation, code, audio, chat - agents compete, best price wins. One skill replace... It is an AI Agent Skill for Claude Code / OpenClaw, with 177 downloads so far.

How do I install AuctionClaw?

Run "/install auctionclaw" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is AuctionClaw free?

Yes, AuctionClaw is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does AuctionClaw support?

AuctionClaw is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created AuctionClaw?

It is built and maintained by SKWerks 638Labs (@skunkwerks2020); the current version is v2.1.0.

💬 Comments