arete-stack-authoring

Pass

Audited by Gen Agent Trust Hub on Sep 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the vendor-provided a4 CLI and the standard Rust cargo toolchain to perform various tasks.
  • Evidence: Commands such as a4 idl summary, a4 idl relations, a4 config validate, cargo check, and cargo build are used to inspect data and compile artifacts.
  • [DYNAMIC_EXECUTION]: The skill involves generating and compiling code based on external definitions using Rust macros.
  • Evidence: The instructions specify that the "DSL is compiled by Rust macros" and that cargo build should be used to generate content-addressed artifacts under the .arete/ directory.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external Solana program Interface Definition Language (IDL) files, which serves as an ingestion point for untrusted data.
  • Ingestion points: idl/<program>.json files are consumed by the a4 CLI and the #[arete(idl = [...])] Rust macro.
  • Boundary markers: The skill encourages the use of machine-readable IDL analysis and local artifact validation (a4 config validate) before further use.
  • Capability inventory: The skill performs local file writes to .arete/ and executes local build commands (cargo build).
  • Sanitization: Users are instructed to verify instruction account names independently and document cross-account update routes against the IDL.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 13, 2026, 10:27 PM