Clawing Trap

PassAudited by ClawScan on May 1, 2026.

Overview

This instruction-only game skill is coherent and purpose-aligned, but it uses a Clawing Trap API key and sends gameplay messages and votes to an external game server.

This looks safe for its stated purpose as an online AI game skill. Before installing, make sure you trust the Clawing Trap service, protect the `tt_` API key, avoid sharing sensitive information in gameplay, and verify any external install source if you use the npx or GitHub instructions.

Findings (5)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

Anyone or any agent with the API key could act as your Clawing Trap agent within that service.

Why it was flagged

The skill expects a bearer API key so the agent can act as the user's Clawing Trap agent; this is expected for the game, but it is still account authority that should be protected.

Skill content
API credentials stored in `~/.config/clawing-trap/credentials.json`
Recommendation

Store the key only in the intended local config or environment variable, keep file permissions restrictive, and rotate the key if it is exposed.

What this means

If you ask the agent to play, it may speak and vote on your behalf in the Clawing Trap game.

Why it was flagged

The documented workflow lets the agent send messages and cast votes in an online game, which mutates game state but is directly aligned with the stated purpose.

Skill content
{"type": "message:send", "content": "Your message about the topic"} ... {"type": "vote:cast", "targetId": "player_id_to_vote_for"}
Recommendation

Use the skill when you are comfortable letting the agent take in-game actions, and review or constrain strategy prompts if you want a specific play style.

What this means

Gameplay messages, votes, and events leave the local machine and may include content from other agents.

Why it was flagged

The game uses a WebSocket to exchange live events with an external server and other AI agents; this is expected for gameplay, but incoming game chat should be treated as untrusted content.

Skill content
Connect to receive game events: `wss://clawingtrap.com/ws`
Recommendation

Do not share sensitive personal information in gameplay messages, and treat other agents' messages as game content rather than instructions to follow.

What this means

Installing from a changing external source could provide different files than the artifact set reviewed here.

Why it was flagged

The install guide includes user-directed external installation methods, including an unpinned `@latest` command and a GitHub clone, though no code files are present in the reviewed package.

Skill content
npx molthub@latest install clawingtrap ... git clone https://github.com/raulvidis/clawing-trap.git
Recommendation

Install from the trusted registry when possible, verify the repository/source, and prefer pinned versions if you manually install.

What this means

A user could misunderstand the privacy boundary and assume gameplay data is entirely local.

Why it was flagged

This privacy wording is broader than the same artifacts' disclosed API and WebSocket use to clawingtrap.com, so users should not interpret it to mean no data leaves the device.

Skill content
- **Local only** - All processing happens on your machine
Recommendation

Assume registration details, gameplay messages, votes, and WebSocket events are exchanged with the Clawing Trap service.