v3.61.0 — ruRE reads CUDA (rure_analyze_cuda)

Minor release: @claude-flow/cli, claude-flow, ruflo 3.61.0. Adds one MCP tool (rure_analyze_cuda). No leaf package changed, so no standalone publishes.

Added

• ruRE: CUDA static analysis (ruflo-rure 0.5.0, plugin ADR-0006). New read-only MCP tool rure_analyze_cuda reads the GPU code inside a CUDA fatbin, cubin or PTX file: fatbin entries (PTX or cubin, target architecture, LZ4/zstd compression), each kernel with register count, code size and constant/shared/local memory, and PTX version, target and kernel names. It parses only; nothing is executed, and strings taken from the file pass the AIDefence gate. Checked against real nvcc 12.8 and 13.0 output. Not included: SASS decoding, cuobjdump shell-out. zstd entries need Node 22.15+.
• ruflo-console 0.47.0: a CUDA row on the ruRE page, DETECT routes for .cubin, .fatbin and .ptx, and a hint on a native binary that holds a CUDA fatbin (ADR-495..497). The ruRE page refuses a sample that is a symlink, wraps long lists, and shows Claude a structured rure block in console_state.
• ruflo-mods 0.7.0: opt-in rureOffer (off by default, ADR-498) adds one line under an answer when the model touched a binary, library, .asar or CUDA file, saying ruRE can inspect it without running it. Offer only; the mods plugin is not part of the CLI tarball.

Fixed

• Dependency audit: zod 4.6.7 in ruflo-graph-intelligence and @claude-flow/browser (GHSA-mrhx-c7xc-m3w5). Nested zod 3 copies stay under two time-boxed accepted findings that expire 2026-11-23 (#4442, #4444).