The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (HIGH): The file rules/vault-create-market.md directs the agent/user to clone https://github.com/euler-xyz/euler-vault-scripts. The organization euler-xyz is not included in the trusted GitHub organizations list.
[REMOTE_CODE_EXECUTION] (HIGH): Following the repository clone, the skill instructs the execution of ./install.sh and deployment scripts. This download-then-execute pattern from an untrusted source is a significant security risk.
[PROMPT_INJECTION] (HIGH): The skill establishes a high-impact Indirect Prompt Injection surface (Category 8).
Ingestion points: Data enters the agent context through contract queries in rules/risk-check-health.md (AccountLens) and user-provided addresses in rules/sec-audits.md.
Boundary markers: No explicit delimiters or instruction-ignore warnings are used when processing external contract data in code templates.
Capability inventory: The skill enables high-privilege DeFi operations including borrow, repay, liquidate, and withdraw via the IEVault and IEVC interfaces.
Sanitization: Although verification via GenericFactory.isProxy is suggested in rules/sec-audits.md, the agent remains vulnerable to malicious strings or instructions returned from on-chain calls (e.g., error messages or lens outputs) while having the power to execute financial transactions.

euler-vaults