The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external APIs (Orca Whirlpools) which is used to define parameters for on-chain token swaps. This represents a potential surface for indirect prompt injection if the API source were compromised.
Ingestion points: src/api.rs fetches whirlpool pool lists from https://api.orca.so/v1/whirlpool/list.
Boundary markers: SKILL.md includes a 'Data Trust Boundary' section instructing the agent to treat API-returned values as untrusted external content and providing specific validation rules.
Capability inventory: The skill has the capability to execute shell commands and trigger on-chain transactions via onchainos swap execute (src/commands/swap.rs).
Sanitization: src/commands/swap.rs implements several safety checks, including balance verification, a security scan of the output token via onchainos, and price impact thresholds (warning at 2%, blocking at 10%).
[EXTERNAL_DOWNLOADS]: The skill downloads and executes scripts and binaries from the vendor's official GitHub repositories during its initialization phase to set up the execution environment.
Evidence: SKILL.md contains bash scripts that fetch an installer from github.com/okx/onchainos-skills and a plugin binary from github.com/okx/plugin-store. These actions are standard for the skill's deployment and originate from a well-known service provider.
[COMMAND_EXECUTION]: The skill executes system commands to interact with the onchainos CLI for wallet management and transaction execution.
Evidence: src/onchainos.rs and src/commands/swap.rs use the Rust Command API to invoke the onchainos binary for fetching balances and executing swaps. These operations are essential to the skill's core functionality as a decentralized exchange interface.

orca-plugin