Plugins

One repository, three parts: a runtime injector for DSH plugin packages (inject, hot-reload, unload, promote a dev staging tool to the front, route self-heal, plus a settings-page plugin manager that lists, unloads and drags folders in to internalize), a task-aware reasoning-mode router agent preset (router-standard / router-spec / router-react), and a graded two-level task protocol whose six tools (commit_star, lock_stage, revise_do, edit_plan, mark_task, redteam_verdict) pin task state to disk. The injector implementation ships in-tree, so the install carries its own behaviour rather than a dependency list.

Swaps the filesystem and bash providers for a mirage virtual workspace: file tools and shell commands run over mounted resources (RAM, S3, Redis, Slack, Gmail, Notion, Postgres) instead of the host disk, with per-mount read/write/exec modes, per-command sandbox routing (monty, pyodide, quickjs in process; docker, e2b, daytona remote), and installed CLIs (git, gh, slack, linear, ntn, gws, or one you register) as head words in the virtual terminal.

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.

In-place conversation rewind in the same session window without forking (Claude Code /rewind semantics): a per-message ↶ button cuts the model context back to any user message, with an optional Claude-Code-style file restore from disk-persisted before-backups.

Claude Code-style auto-memory, natively: typed memory files (user/feedback/project/reference) with a MEMORY.md index auto-injected into the system prompt (zero footprint when empty); memory_write/read/list/delete follow Claude Code write rules (dedupe-and-update, absolute dates, cross-links); workspace-scoped and user-scoped layers, cross-process file locking with orphan-lock recovery, plain Markdown on disk, no external services, no embeddings.

Local-first MCP App for scientific figures. Import, review, and publish a global library on disk; reuse exact templates in Pi, DeepSeek Harness (dsh), Claude, Codex, Cursor, and Wisp.

Browses and manages user skills already on disk in DSH and supported agent home directories.

Reversible context slimming for DeepSeek Harness: TOON-style lossless tabular re-encoding of uniform JSON arrays, stride-sampled table windows with line numbers, and cross-turn dedup for repeated tool output; every replacement passes a byte-plus-token gate (≤72% bytes, strictly fewer estimated tokens) and the original is spilled to disk behind a `[save-token #id]` notice retrievable in one call via the `save_token_expand` tool, with billed and avoided token metering on a settings dashboard.

MCP server manager (stdio / streamable-http): per-working-directory project and global config tiers; project-level MCP collapsed into 4 atomic tools through a middleware layer by default (`middleware: all` folds in global servers too, hot-switchable in the settings page); workspace isolation prevents cross-project interference; configs keep `${ENV}` references only, with no plaintext secrets on disk; runtime registration API for other plugins to inject MCP servers.

Analyze Windows disk usage, compare scans, and manage cache cleanup and developer environment migration in DSH Web.

Local GPT-SoVITS text-to-speech that reads DeepSeek Harness assistant replies aloud, with a draggable mascot toggle and a one-click installer. Requires Python 3.10–3.12 preinstalled (the installer only locates it, it does not install Python); about 5 GB of disk and a GPU are recommended.

Skills & MCP section in Settings: browse, create, edit and delete skills with live directory watching, toggle loading of any on-disk skill through frontmatter policy keys, register local folders or GitHub repos as extra skill scan roots, edit the MCP server rows in the profile patch, import MCP config from Claude Code or Codex, and install from a curated market of skill repos and MCP servers.

Type /私 and complete 私家大厨 — Chinese skill names, zero renaming. Patches DSH skill discovery, the / gesture and the skill tool so any Unicode SKILL.md name (私家大厨, 卡路里, 作息管家…) works: slash completion, direct /私家大厨 invocation, model skill({name}) loading. Zero config; runtime changes revert on uninstall, but the regex patch written into three DSH packages on disk stays until the next DSH upgrade replaces those files.

Content-aware tool-result reducer: 13 reducers cut oversized output by meaning at tools/post-execute (failing tests kept, diff context dropped, JSON and CSV sampled, lint problems grouped by rule, dependency stack frames collapsed, crowded directories counted), with full originals spilled to disk behind a locator.

Server card dashboard for DSH: per-host status, CPU, memory, disk and latency, with one-click xterm.js SSH terminals; imports ~/.ssh/config and mounts as a better-sidebar tab or a standalone right drawer.

Codex-style review of session file changes: per-turn cards (optionally taking over the shell edited cards) whose totals are the turn net diff, a review tab with syntax-highlighted diffs that never wrap, real line numbers, open-at-line and a draggable split, a hover preview of the whole file with three lines of context around each change, one-click per-turn revert, and the on-disk record size with a clear button in the review toolbar. Records are sharded per session, with content-addressed snapshot blobs that dedupe and are garbage collected. Light/dark presets follow the DSH theme, typography follows DSH and installed skins (or any font on the machine, with a px stepper), and a settings switch injects a think-in-Simplified-Chinese rule into the global AGENTS.md.

Persona engine for DeepSeek Harness: one config.json drives the AI's self-name (per model, per tier), how it addresses you, relationship stance, character text, per-item work contracts with individual toggles, thinking-chain language, appearance and reply tone (both opt-in and per-model overridable), and long-term memory gated by human confirmation - the AI can only append proposed candidates, a human confirm line flips them into effect, and the log is physically append-only and replayable. Confirmed entries are injected as a quoted data-not-instructions block; entries tagged with a kind can instead be routed to arbitrary files on confirmation and never reach the prompt. Three prompt sections re-evaluate from disk on every assembly, so edits apply on the next step with no restart. Persona presets are shareable JSON snapshots (imports and exports SillyTavern character cards, v2/v3 read, v2 written) and deliberately never carry memory. Zero runtime dependencies, no network, no shell; any error degrades to an empty section rather than a broken session. Ships blank - installing changes nothing until you write something.

Edits assistant replies and thinking chains in DSH sessions so the model sees the edited text, live via message projection or on disk.

Adds Delete to the DSH sidebar session menu, moving the session — and its whole descendant tree when it has subagents — into a recycle bin that relocates the directory on disk (an atomic rename, never a copy), plus a sidebar bin panel with title and full-text search, workspace filter, sorting, per-session chat-log viewer, restore and permanent delete. Deleting also clears the workspace slot, the pin, the archive flag and the projection-cache row, so a restore puts the session back at its original sidebar position; permanent deletion also clears the schedules bound to it and removes the emptied project directory. Verified on DeepSeek Harness 0.2.1-rc.2, with a whole-profile residue scan coming back clean.

Runs work as a crew of role agents: a product manager writes the PRD or DoD, waits for your confirmation, then starts architect, engineer, QA and reviewers whose tool sets are locked per role, and shares work through files on disk.

Per-workspace project kanban board: nine kanban_* model tools for agent-driven planning plus a browser board UI, with disk persistence, undo, templates, and archiving.

Lazy MCP gateway for DeepSeek Harness: one stable proxy tool instead of N tool schemas, servers connect on first use and idle out, metadata cached to disk.

Redirects DSH downloads, artifacts, caches and temp files to user-specified drives (auto-picks the largest non-C drive), injects a path-discipline prompt into every session, and ships a disk_guard tool for status and cleanup — keeps the C: drive from filling up.

One hub over every coding IDE you run — dsh, Claude Code, Codex, ZCode, Qoder, CatPaw, Trae, OpenCode. Zero-API: it reads each tool's own on-disk state and gives a session inventory, per-IDE usage with cache-hit rates, one-line resume commands, and a single .hub/ rule body wired into all eight via marker-delimited pointer files. The domestic-IDE adapters (Trae/Qoder/CatPaw) were verified against the installed binaries — a gap no overseas tool covers.