The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation instructs the installation of external dependencies during the setup phase.
Evidence: npm install starknet@^8.9.1 @avnu/avnu-sdk@^4.0.1 in SKILL.md.
These are well-known packages in the Starknet ecosystem provided by established developers.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform environment setup and dependency management.
Evidence: Use of npm install for package management in the prerequisites section.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from the AVNU API and protocol sources which could potentially influence agent behavior or transaction parameters.
Ingestion points: The skill fetches market data, quotes, and order statuses through the AVNU SDK via methods like getQuotes, getPrices, fetchTokens, and getDcaOrders in SKILL.md.
Boundary markers: No explicit delimiters or boundary markers for external data are defined in the instructions or prompts.
Capability inventory: The skill possesses significant capabilities including executing blockchain transactions (executeSwap, executeCreateDca, executeStake, executeClaimRewards, executeUnstake), file operations (Read, Write), and shell access (Bash).
Sanitization: The provided code snippets demonstrate direct use of API-returned quote objects in execution functions without explicit validation or sanitization layers beyond standard SDK usage.

starknet-defi