Prompt optimization that cuts token cost — system roles, stop sequences, word caps, no preamble. Works with Kimi Code, Codex, and Claude Code.

Save, search, and recall notes in the shared vault that both the Chat tab and Code tab read from — continuity across sessions, no re-deriving context.

Compresses notes before they hit the Claude-Desktop-Bridge vault, so shared memory between Chat and Code stays cheap to write and recall.

Builds smart wordlists ranked by how real people actually think, and cleans existing lists for hashcat/aircrack workflows.

Fact-checks AI-generated research reports claim-by-claim against primary sources. Errors first, compliments last.

Moves, copies, renames, and organizes files — never deletes without asking, never runs sudo without approval.

Ultra-compact communication mode — cuts roughly 50-65% of output tokens while keeping every technical detail intact.

Compacts a conversation into a handoff document — decisions, work done, suggested next skills — so a fresh agent can pick up cold.

Controls Audacity through mod-script-pipe and teaches mixing/mastering technique calibrated to your genre.