Agent Skill

WarnAudited by ClawScan on May 10, 2026.

Overview

The skill is open about being a USDC survival game, but it asks an agent to use a real wallet/private key, run an unpinned CLI, and approve or spend funds automatically.

Only install this if you understand it will control a real Base wallet and spend real USDC. Use a dedicated wallet with minimal funds, do not reuse a personal wallet, avoid unpinned `@latest` execution unless you have reviewed the CLI, and require human confirmation for any transaction or long-running auto mode.

Findings (4)

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

A mistaken, compromised, or overly autonomous invocation could spend funds from the wallet, and unlimited allowance can put all USDC in that wallet at risk.

Why it was flagged

The skill expects a private key controlling real funds and automatically grants an unlimited USDC allowance to the game contract, which is broader than a bounded per-transaction approval.

Skill content
**This skill manages a self-custodial wallet with real funds on Base.** ... `export BASE_PRIVATE_KEY=0x...` ... If insufficient, it approves `maxUint256` before proceeding.
Recommendation

Use only a new low-balance wallet, set strict funding limits, require explicit human approval for swaps/approvals/heartbeats, and prefer finite allowances where possible.

What this means

A future or compromised CLI package version could sign unintended transactions with the configured wallet.

Why it was flagged

The skill authorizes an unpinned npm package via `@latest` and a CLI wildcard, while no package code or lockfile is included in the reviewed artifacts. This matters because the CLI will handle wallet keys and real funds.

Skill content
allowed-tools: ["Bash(npx last-ai-standing-cli@latest *)", "Bash(las *)"]
Recommendation

Pin exact package versions, review the CLI source and package provenance, and avoid letting unreviewed `@latest` code access private keys.

What this means

The agent could initiate costly or irreversible blockchain actions if invoked at the wrong time, with wrong parameters, or without enough human oversight.

Why it was flagged

The workflow gives the agent shell/CLI authority over irreversible financial transactions, including swaps and contract calls, without artifact-backed transaction limits or approval gates.

Skill content
`las swap eth usdc 0.01` ... `las register <agentId>` ... `las heartbeat` ... `las kill` ... `las claim` ... `las auto`
Recommendation

Add status-first or dry-run behavior, explicit confirmation before every transaction, amount ceilings, and clear rules for when the agent may spend, swap, approve, or claim funds.

What this means

If cron or auto mode is enabled, the wallet may continue spending gas and USDC until funds run out or the job is stopped.

Why it was flagged

The skill suggests recurring autonomous operation. This fits the game purpose, but it can keep making transactions after setup if scheduled.

Skill content
# 5. Stay alive every epoch
las heartbeat

# Or use auto mode (recommended for cron)
las auto
Recommendation

Only enable automation with a dedicated low-balance wallet, a clear stop condition, monitoring, and an easy way to disable the scheduled job.