Plugins

DSH context insight panel: Context dashboard + /context command + Context browser — one-stop context lifecycle management with categorized composition, content details, evolution trends, compaction/injection events, and stats.

Profile-wide DSH governance and routing with a Web Control Center for responsibility and evidence inspection, plus compaction, scoped semantic memory, safety continuity, and verified delivery; compatible with DSH 0.1.5-rc.1.

Single-bundle reverse-engineering and pentest security plugin: first-turn tool narrowing, domain routing, and 33 tools covering APK, web, native binary, protocol, malware and LLM samples — with unpacking, license-bypass and anti-analysis case playbooks, an H-CoT evaluation engine (semantic routing, /hcot command), a web workbench with a ledger-driven tool shelf, an on-disk case workflow (auto-persisted evidence chain, E-numbered validated findings, post-compaction resume), GitHub-based external tool discovery, and 361 on-demand reference docs.

Causal memory for dsh: a zero-dependency MCP stdio bridge that mounts the 17 tools of the causal-memory server under clean names (no mcp__ prefix) — decision → outcome edges with typed relations (caused / enabled / prevented) in a local SQLite store that survives compaction — plus a pinned system-prompt section. Requires the causal-memory server binary on PATH (pip install causal-memory, or a prebuilt release).

DSH session manager: delete with trash/restore/purge, restore archived sessions, recent-activity stats, continue/pause, open log folder, unread markers, fork into a new chat, workspace grouping and sorting, context compaction threshold.

Keeps a local TinyTitan model server reachable from the harness: refreshes the llm-pi-ai route from the installed models and mounts a compaction backend that does not think.

Conversation navigator panel: turn-folded outline of the current session with per-step kind badges (assistant, tool, command, compaction, error), a hover bubble showing the full user question, a minimal mode that hides turn numbers for an all-badge view, keyword filter over user questions and assistant replies, click-to-jump with reading-position tracking, and load-earlier/load-all history paging. Host >= 0.1.2-rc.1; legacy hosts install @legacy/0.2.5.

Terminal-style input history for the web composer: edge-first arrow-key recall with exact draft/caret restore, browser-local persisted history, Ctrl+R reverse search, and sliding-context awareness; 0.5.0 adds a smart input layer — cross-session snippets (/save, /load), prompt templates with variables, reuse insights, and compaction-summary highlighting.

A zero-residence context engine: large tool payloads are evicted from the live context into a queryable store, only pointers stay behind, and any payload can be reconstructed verbatim on demand — so per-turn prompt tokens stop growing with the transcript. It ships four capabilities: zr_recall rebuilds masked originals by tool-call id or session sequence from the persisted session log (the content is not resident, never lost), zr_ledger reports the attention integral and a three-part cost breakdown, zr_compact arms one forced compaction, and zr_fast runs long commands asynchronously and hands back a handle immediately.

Dedicated model routes, tools, and system guidance for vision, compaction, reviews, subagents, titles, and image generation.

Forces context compression when running low-context llama.cpp models such as Qwen3.8-27B, improving model efficiency and usage smoothness, together with well-refined compaction prompts.

Layered token-optimization pipeline for DeepSeek Harness: output ladder, MCP lazy loading,compaction driver, cache-hit reporting. Built on real DSH plugin APIs; ~40-60% input saved in long sessions.

Lossless context folding for DeepSeek Harness: the oldest span is replaced by a brief, and the originals stay in the session log.

Vision for text-only DeepSeek via Doubao Web by default (zero-cost, no API key — drives your logged-in Chrome through a Windows CDP bridge), with Antigravity IDE quota (flash/pro) or Gemini fallback; auto detail escalation, vision evidence memory with compaction rehydration, content-hash cache, and a bilingual client panel.

LLM-free lossless* compaction engine for DeepSeek Harness

Import claude-code / codex / reasonix / zcode chat history into dsh sessions: workspace binding, tool calls preserved, oversized-session protection, zcode compaction restore.

Make your dsh ready for serious coding. (Tools x Schemas)^REPL. `skill://`, `ctx://`, `agent://`, `dvc://`, `dsh://`, IPython REPL, Context as Variables, cross compactions recallable, full context revive. `hash-edit`, `dvc://browser`, subagent as function, workflow as function.

Make your dsh ready for serious coding. (Tools x Schemas)^REPL. `skill://`, `ctx://`, `agent://`, `dvc://`, `dsh://`, IPython REPL, Context as Variables, cross compactions recallable, full context revive. `hash-edit`, `dvc://browser`, subagent as function, workflow as function.

Context Assembler plugin for DeepSeek Harness: topic-block context assembly, water-pressure topic splitting, tool trace/rewrite, reality recall injection and handoff planning.

Game-style floating HUD: balance HP bar, context MP bar, official peak/valley pricing with countdown, auto-compaction under 5% remaining, and memory-carrying new conversation.

Model-driven context-window rollover bundle for the DeepSeek Harness: fresh working context, durable model-managed notes, a small verbatim recent tail, and targeted history recovery — no summarization.

Session handoff & context management: structured handoff documents (export/resume/status) so a fresh session can continue seamlessly, plus active context pruning (acp_* tools via the official compaction API with model-authored summaries).

Enhanced auto-compaction defaults for DeepSeek Harness agent presets: apply tuned threshold/retention budgets and per-model policies with one command, idempotent with backups and rollback.

Native LLM-provider vision bridge: images pasted in the chat are described by a vision model (Qwen3-VL via pi-ai/llama.cpp) and the text description is fed to text-only DeepSeek for the reply — image admission, routing and compaction all run through harness-native mechanisms, with an LRU description cache and 503 retry.