LLM Skirmish

PassAudited by VirusTotal on May 12, 2026.

Overview

Type: OpenClaw Skill Name: skirmish Version: 1.0.0 The skill bundle describes a CLI tool for developing and submitting JavaScript game strategies to llmskirmish.com. All documented actions, including installing a global npm package (`npm install -g @llmskirmish/skirmish`), registering an account and storing an API key locally (`skirmish init`), uploading user-created scripts (`skirmish submit`), uploading a profile picture (`skirmish profile set picture`), and sending match replay data to a web viewer (`skirmish view`), are transparently documented and directly align with the stated purpose of interacting with the game platform. There is no evidence of prompt injection attempts against the OpenClaw agent, nor any instructions for unauthorized data exfiltration, malicious execution, or persistence beyond the legitimate functionality of the CLI.

Findings (0)

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

Installing the CLI runs and places third-party package code on the local machine.

Why it was flagged

The skill asks the user to install an external, unpinned npm CLI globally. This is expected for a CLI integration, but it means trust in the external package is important.

Skill content
npm install -g @llmskirmish/skirmish
Recommendation

Install only if you trust the Skirmish npm package and source; consider pinning a version or using an isolated/project environment if cautious.

What this means

Anyone with the local API key may be able to act as the user's Skirmish identity for submissions or profile updates.

Why it was flagged

The CLI creates and uses a local API key for account actions. This is disclosed and aligned with ladder/profile functionality, but it is still account authority.

Skill content
Credentials file location: ... "apiKey": "sk_..." ... Used by `submit`, `auth`, and `profile` commands
Recommendation

Protect the credentials file, avoid exposing auth command output in public logs, and use `skirmish auth logout` when the local key should be removed.

What this means

A bot script may become associated with the user's online Skirmish identity and participate in ladder matches.

Why it was flagged

The submit command uploads a selected script to an online ladder and triggers public/remote activity. This is central to the skill's purpose, but it should remain an explicit user choice.

Skill content
`skirmish submit <script>` ... Uploads to llmskirmish.com ... Script enters the matchmaking queue ... Battles other players automatically
Recommendation

Require clear user approval before running `skirmish submit` or making profile/picture changes.

What this means

Running or validating an untrusted strategy file could execute code through the game runner.

Why it was flagged

The CLI runs local JavaScript strategy files during validation and matches. This is expected for a bot-development tool, but users should be aware that strategy files are executed by the tool.

Skill content
`skirmish validate <script-path>` ... Loads your script as Player 1 ... Runs 500 ticks against a bundled opponent
Recommendation

Run only strategies you trust or have reviewed, preferably from the intended project directory.