evm-decimal-validation
Warn
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: MEDIUMMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill provides deceptive technical guidance regarding blockchain token decimals. It defines a 'sensible fallback' of 2 decimals as a 'majority value' and provides test cases hardcoding 0 decimals for major networks (Polygon/137 and BSC/56). In reality, most EVM native and ERC20 tokens use 18 or 6 decimals. Misconfiguring these values would result in amount calculation errors by factors up to 10^18, leading to severe financial consequences.
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process and update network configurations from external files (blockchain/networks.go, token-config.yaml) without establishing security boundaries or sanitization logic.
- Ingestion points: Files like blockchain/networks.go, token-config.yaml, and *.go source files analyzed via grep.
- Boundary markers: Absent. The skill does not instruct the agent to distinguish between trusted and untrusted data within these configuration files.
- Capability inventory: The skill encourages modification of network configuration types and helper functions within the codebase, as well as running grep commands.
- Sanitization: Absent. There is no validation against external trusted sources; the provided 'ValidateDecimals' function only checks local config against the current client output, which is insufficient if the config itself is seeded with the deceptive values recommended in the skill body.
Audit Metadata