Release v1.10.18

Fixed

False Positive Reduction v4 (10 detectors, 52 FPs eliminated, net -22)

Ground truth validation: 311 → 289 findings across 18 test contracts (7.1% reduction), 0 TP regressions.
Cumulative from v1.10.14: 436 → 289 (34% total reduction across 30 detectors in 4 rounds).

• delegatecall-to-self (round 3): Fix self-match bug in selector pattern check — function body includes declaration line causing trivial match; strip declaration before checking
• price-manipulation-frontrun: Skip view/pure functions, flash loan providers, governance contracts, ERC-4626 vaults, and access-controlled functions
• governance-parameter-bypass: Skip state variable declarations, access-controlled functions, governance mechanisms (timelock/require/proposal), and interface definitions
• fallback-function-shadowing: Contract-scoped is_proxy_contract and has_if_admin_pattern (not file-level); detect receive-to-fallback delegation patterns
• aa-social-recovery: Contract-scoped source extraction; skip non-social-recovery contracts (paymasters, nonce managers, session key handlers)
• unprotected-initializer: Require upgradeable context (proxy indicators); detect initialization guards, inline access control, simple ownership setters; strict name matching
• storage-collision (round 3): Gate variable-target findings behind has_proxy_storage_pattern check — require proxy storage patterns before flagging variable collision
• missing-chainid-validation: Multi-layered bridge classification (same approach as bridge-message-verification); access control detection; EIP-712 domain detection; bare receive exclusion
• dos-unbounded-storage (round 3): Skip mapping declarations; require push operations for array findings; LHS-only nested mapping writes; boolean authorization patterns
• inefficient-storage: Word-boundary matching for storage variable names; skip view/pure, flash loan, and external call functions; skip governance parameters from constant check

Note: 36 findings from related detectors surfaced due to cross-detector deduplication effects (when suppressing one detector's findings, similar findings from overlapping detectors become visible).