standards

Warn

Audited by Gen Agent Trust Hub on Oct 28, 2026

Risk Level: MEDIUMMETADATA_POISONINGPROMPT_INJECTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [METADATA_POISONING]: The skill contains deceptive claims regarding the deployment status of various Ethereum standards. It asserts that ERC-8004 was deployed on January 29, 2026, and that EIP-7702 was shipped in May 2025, which represents future-dated or fictional information intended to mislead the agent about the current state of technology.
  • [PROMPT_INJECTION]: The content uses authoritative override language designed to bypass the agent's internal training. Specifically, the phrase "You almost certainly don't have this in training data" is a pattern used to force an LLM to prioritize the provided text over its own established knowledge.
  • [EXTERNAL_DOWNLOADS]: The skill references and provides installation instructions for several unverified third-party packages and repositories, including @x402/express, @x402/fetch, @x402/evm, and @x402/core. These packages are associated with a purported "x402" protocol that lacks current widespread recognition or verification in official registries.
  • [INDIRECT_PROMPT_INJECTION]: The skill creates a vulnerability surface by presenting fictional standards as technical documentation that an agent might use for future decision-making.
  • Ingestion points: The entirety of SKILL.md is ingested as authoritative technical standards documentation.
  • Boundary markers: None identified; the instructions are presented as direct facts.
  • Capability inventory: The skill includes code snippets for x402Fetch and ethers.Contract calls which involve network requests and transaction signing capabilities.
  • Sanitization: None; the agent is expected to trust the provided contract addresses (e.g., 0x8004...) and endpoints (e.g., 8004.org) without verification.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 28, 2026, 02:26 AM