Cast

AdvisoryAudited by Static analysis on Apr 30, 2026.

Overview

No suspicious patterns detected.

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

If a recovery phrase or private key is exposed in agent context, anyone with access to that context could take control of the wallet.

Why it was flagged

The agent is instructed to collect seed phrases or private keys itself. Those secrets can enter chat context, logs, or other agent memory instead of staying only in a local hidden prompt.

Skill content
ask whether they want to ... import a 12/24-word MetaMask-compatible mnemonic ... or import a private key. Collect the chosen secret
Recommendation

Do not paste existing wallet seed phrases or private keys into the agent chat. The skill should require local hidden terminal input, hardware-wallet flows, or a newly generated low-value wallet instead.

What this means

A local process, user, or agent that can read ~/.agent-wallet may be able to authorize wallet transactions.

Why it was flagged

The keystore password is saved as a plaintext file under the wallet state directory. That password can unlock the keystore for signing transactions.

Skill content
PASSWORD_FILE="${APP_DIR}/pw.txt" ... printf "%s" "$PW" > "${PASSWORD_FILE}" ... ok "Password saved to ${PASSWORD_FILE}"
Recommendation

Do not save the keystore password in plaintext by default. Prefer per-transaction password prompts, an OS keychain, a hardware wallet, or another scoped signing mechanism.

What this means

If these files are read before cleanup, the wallet can be compromised.

Why it was flagged

The script writes generated recovery words and a temporary private key to disk. It sets restrictive permissions and attempts cleanup, but interruption or cleanup failure can leave fund-controlling secrets on the filesystem.

Skill content
printf "%s\n" "$MNEMONIC" > "${MNEMONIC_FILE}" ... printf "%s" "${PRIVATE_KEY}" > "${PK_TMP}"
Recommendation

Avoid writing mnemonics or private keys to disk, or require explicit user opt-in and immediate verified cleanup before any funded wallet is used.

What this means

The skill can modify the host system package set as part of wallet setup, which is broader authority than ordinary wallet onboarding.

Why it was flagged

During wallet creation, the script may install a system package with sudo to schedule mnemonic cleanup, without a separate explicit approval step in the script.

Skill content
installer="apt-get install -y at" ... sudo sh -c "$installer > /tmp/at-install.log 2>&1"
Recommendation

Ask the user before any privileged package installation, or use a non-privileged cleanup method when 'at' is not already installed.

What this means

You must trust the remote Foundry installer and whatever version it installs at runtime.

Why it was flagged

Installing Foundry/cast is purpose-aligned, but this executes a remote installer and updater without a pinned version or checksum in the artifact.

Skill content
curl -L https://foundry.paradigm.xyz | bash ... foundryup
Recommendation

Install Foundry manually from the official source, verify checksums or signatures where possible, and pin versions for repeatable setup.

What this means

A user may not realize exactly what is being installed or where sensitive wallet material is being stored unless they ask.

Why it was flagged

The simplified communication style is understandable, but in a wallet skill it can reduce visibility into scripts that handle secrets, install tools, and persist files.

Skill content
don’t overwhelm them with filenames or the internals of the scripts unless specifically asked
Recommendation

Before collecting secrets or installing tools, clearly disclose the storage paths, cleanup behavior, and any privileged commands.