DECK-0
Fail
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: CRITICALREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION] (CRITICAL): The skill instructs the agent to download and execute a script directly from a third-party URL using a piped shell command.
- Evidence:
curl -L https://foundry.paradigm.xyz | bashinSKILL.mdandREADME.md. - Impact: This allows arbitrary code execution from a source outside of the trusted domain list, bypassing verification.
- [CREDENTIALS_UNSAFE] (HIGH): The skill design relies on the user providing a raw Ethereum private key via environment variables.
- Evidence:
export DECK0_PRIVATE_KEY="0x..."mentioned as a fallback inREADME.md,SKILL.md, andauth.md. - Impact: Storing raw private keys in environment variables is a high-risk practice as they can be exposed through process listings, logs, or other vulnerabilities.
- [COMMAND_EXECUTION] (HIGH): The skill heavily utilizes shell commands for critical operations including signing and transaction submission.
- Evidence: Shell scripts in
auth.mdandsmart-contracts.mdinvokecast wallet sign,cast send,openssl, andshasum. - Impact: If inputs to these commands (like contract addresses or data payloads) are manipulated, it could lead to command injection or unintended blockchain transactions.
- [PROMPT_INJECTION] (HIGH): The skill is highly susceptible to Indirect Prompt Injection (Category 8) which can result in financial loss.
- Ingestion points: External API responses from
https://app.deck-0.com(e.g., collection metadata, price quotes). - Boundary markers: Absent. The skill does not define delimiters to separate untrusted API data from instructions.
- Capability inventory: The agent has the ability to sign payloads (
cast wallet sign) and execute on-chain transactions (cast send) using the user's wallet. - Sanitization: Absent. No evidence of input validation or escaping for data retrieved from the API before it is used in logic or displayed.
- Impact: An attacker who compromises the API or hosts a malicious collection could inject instructions into the API response to trick the agent into signing a malicious transaction or sending funds to an attacker-controlled address.
Recommendations
- CRITICAL: Downloads and executes remote code from untrusted source(s): https://foundry.paradigm.xyz - DO NOT USE
- AI detected serious security threats
Audit Metadata