solana-defi-architect

Run the audit pass which applies a checklist covering missing signer checks, owner checks, PDA misuse, arbitrary CPI, sysvar spoofing, and account confusion, combined with threat modeling and an impact-times-likelihood risk matrix. Critical and High findings require an executable proof-of-concept test before being reported as confirmed.

It uses a three-tier pyramid: LiteSVM for fast in-process unit tests, Mollusk for isolated single-instruction and compute-budget checks, and Surfpool for integration tests against realistic cluster state. Fuzzing runs via anchor fuzz by default and Trident for release-grade, invariant-driven campaigns.

Yes, but only on explicit request, with confirmation of the upgrade-authority plan first. It supports local keypair files, Ledger hardware wallets, and Squads multisig, and recommends verifiable builds plus multisig upgrade authority for programs holding real value.

Most Anchor build failures trace to version mismatches between anchor-lang in Cargo.toml and the Anchor CLI, or Rust toolchain incompatibilities such as the Anchor 0.30.x plus Rust 1.80 time-crate issue. The skill checks anchor, solana, and rustc versions before writing code to catch this early.

It still writes the code, tests, and scripts, but explicitly labels every result as NOT EXECUTED with the exact local command to run, rather than claiming tools ran. This prevents unvalidated findings from being reported as confirmed in the audit report.

Use it when there is Solana-specific evidence such as .rs files with anchor_lang or solana_program imports, or explicit mentions of Anchor, PDA, or CPI. If evidence points to Solidity or CosmWasm instead, it defers to the sibling chain skill rather than auditing the wrong stack.