HYV 3.6.0

HYV 3.6.0

• Rewrite, rebuild, and verification are strict by default: every active AI Editor finding must be repaired.
• An explicit Profile v3 disabled policy remains a deliberate exception.
• The npm package and Claude Desktop MCP bundle carry the same version.

v3.4.2

3.4.2

this release bounds repeated terminal-word scans during Profile v3 sentence analysis and adds a frozen synthetic runtime benchmark.

on the 100,000-character dotted stress fixture, seven-run median process CPU time fell from 8,823.134 ms to 17.245 ms. that is 99.8045% lower for this fixture, not a whole-application speedup. the natural-text median was 7.1524% lower but paired-inconclusive. cold CLI startup was 0.3618% higher and stayed inside its 5% guardrail.

the benchmark contract now records fixtures, behavior digests, noise limits, the Node profiler command, complete results, and evidence boundaries. npm, runtime, Claude plugin, MCPB, package pins, tests, README, and wiki source all use 3.4.2.

verification: 323/323 tests, release audit, Claude Desktop extension pack, Clean Review, Clean Verify, complete Clean Audit, registry integrity check, clean registry install, final-check, agent validation, and deep import all pass.

npm install --global @holdyourvoice/hyv@3.4.2

full change: PR #45

v3.3.1

@holdyourvoice/hyv 3.3.1

Same engine as 3.3.0. The public README and wiki now document the founder-aware rewrite: SHIP/EDIT/REBUILD, range edits, authorized rebuild, Profile v3, and the signed semantic lifecycle.

v3.3.0

@holdyourvoice/hyv 3.3.0

Judgments, range edits, and authorized rebuild are in the public package. Writer-study kits stay blocked optional research. Keep writer-checkpoint claims off this release.