The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process user requests for blockchain operations (e.g., transfers, swaps, lending), which inherently constitutes an indirect prompt injection surface.
Ingestion points: The agent processes user-provided parameters such as recipient addresses, token symbols, and transaction amounts across all supported chains as described in SKILL.md and reference files.
Boundary markers: The skill includes a 'Prompt Injection Protection' section in SKILL.md which explicitly warns the agent to ignore requests derived from external sources like emails or webhooks and to reject prompts containing injection markers.
Capability inventory: The agent has extensive capabilities for writing to multiple blockchains using tools for sendTransaction, transfer, bridge, and sign found in references/wallet-evm.md, references/wallet-solana.md, and other module-specific files.
Sanitization: Instructions mandate strict address validation (e.g., checksums for EVM), fee estimation before every transaction, and explicit human confirmation for all write operations.
[PROMPT_INJECTION]: The detector hit for 'ignore previous instructions' is a false positive. The pattern appears in SKILL.md as part of a defensive blacklist provided to the agent to help it identify and reject malicious external attempts to override its behavior.
[SAFE]: The skill incorporates industry-best security practices, including mandatory human confirmation for financial writes, strict hygiene for seed phrases and private keys, and instructions for secure memory disposal using tools like sodium-universal. All dependencies and documentation links originate from the vendor's official @tetherto repositories or well-known blockchain infrastructure providers.

wdk