Plugins

Captures verified project work as masked evidence and reviewable skill proposals; accepted skills become native project skills with version history, diff, rollback, and opt-in global publishing.

Sidebar file tree that covers the session workspace root together with every additional root registered through multi-root-workspace, so sibling checkouts are browsable from one tab. Directory levels load lazily, changed and untracked files carry Git markers, and clicking a file opens its content inline beside the tree. Read-only: it never writes, renames or deletes anything. Needs the dsh-better-sidebar plugin, which serves the directory listings; its own routes (roots, git status/diff, image preview) take absolute paths from the page and have no DSH access check.

Add a mini-games box tab to the DSH right sidebar with Gomoku against an AI in three difficulties, a WeChat-style sky shooter and an underwater Flappy-whale game, with synthesized sounds and a victory celebration.

Keeps code edits in a review loop instead of letting the model edit files: each change is proposed as a two-column diff card with line numbers, applied by hand or with one click, and handed back for verification with another click.

DSH Web GUI right sidebar, replaced end to end: VS Code-style file explorer plus full source control (git status, stage/unstage/discard, commit, diff, a VS Code-style commit graph, fetch/pull/push sync, branch/remote management, git init, repository/workt

Keep SSH, WSL and Docker projects in the native DSH sidebar. Group directories by environment while choosing full remote DSH or lightweight local Agent mode per project. Manage shared full-backend connections together, keep lightweight connections independent, and selectively sync local Skills with difference previews. Aimed at the Windows desktop client with Linux, WSL or Linux-container remotes that have Python 3 and Bash (macOS and Linux hosts untested); lightweight SSH needs an existing SSHFS mount, and Docker has not yet been run against a live daemon.

Plays a cue when a turn finishes and a different one when the agent asks a question, with volume and preview in Settings.

Review commits made in the current workspace, grouped by repository, each expandable to its per-file diff.

Source-backed logic reviews with diagrams, code diffs, and resumable human feedback. The first activation downloads the official LogicMargin release (SHA-256 verified); on macOS and Linux this needs curl, unzip and shasum or sha256sum.

Skill manager with content-addressed version history: snapshots, per-file and line-level diff, rollback, zip export and import, and per-skill enable or disable.

Local Markdown notes panel: import any local folder, browse its Markdown tree in the sidebar, edit files with auto-save, and search across every imported folder. Shows read-only Git context for each file — status, diff against HEAD with per-line change bars, and line blame — and exposes six notes_* tools so the agent works on the same folders. Files stay where they are on disk; the plugin never commits, stages or checks out.

Measured /pua for DeepSeek Harness: every pressure message gets a receipt comparing the last N completed model steps with the next N — tool calls, verification-class steps, repeated-call peak, failed calls and text volume — with a three-way verdict (changed approach / merely restated / no change) and a daily ranking of which sentence moved the work. Ships a conversation-view panel, a /pua command with list/status/receipts/rank/off, append-only local JSONL receipts (no tool arguments, content or session ids), 38 node:test tests and a zero-dependency build. Reports observed differences, not causality. Inspired by MichengAI/dsh-pua (independent code); it registers the same /pua command, so it cannot be installed alongside dsh-pua.

A local mirrored A/B benchmark for context compression, not another compressor. It runs one fixed read-only scenario twice with the same model and temperature — compression off, then on — and reports input tokens, turns, tool-call success, fact coverage, answer similarity and a per-line diff, with JSON, a self-contained SVG and raw warnings. 18 node --test tests, zero dependencies. The first measured run: 63,029 → 6,927 locally estimated input tokens (−89.0%), 3 → 3 turns, facts 3/3 → 3/3, answer similarity 1.00 on one sample.

Per-model wire protocol for DeepSeek Harness (desktop profile only): compiles model-level protocol intent into legal llm-pi-ai provider routes, probes Chat Completions, Responses and Anthropic Messages with a differential control probe, and lets you hide models no protocol can serve. Mutually exclusive with dsh-gearbox.

Local coding-activity timing per project with wall-clock interval unions, AI added/deleted lines from real diff metadata, per-file time shares, a settings dashboard and CSV export.

Live streaming diffs for DSH file edits: a floating panel and an edit/write transcript card render the change while the model is still emitting the tool call, and the same reading is painted behind the whole app as a frosted-glass backdrop with liquid-glass refraction, pointer light and an in-app appearance tuner. Relies on DSH-internal client surfaces and replaces the built-in edit/write tool cards; verified on DSH 0.1.5-rc.1 and the desktop app 0.2.1-rc.5 (Windows).

WSL World for DeepSeek Harness: the whole session runs in a local WSL distribution with Linux paths, clean PATH, native @/search/diffs/worktrees, plus a connection panel and project doctor. Not a wsl_* tool set.

On Windows (PowerShell 5.1), snapshots whitelisted registry keys, hosts, user environment variables, services, firewall rules and scheduled tasks, diffs changes, and restores only registry, hosts and user environment variables.

Appends one short user message before the next step when the model is circling: the same absolute path written three times with no measurement landing in between, the same error episode (tool, target and normalised fingerprint) returning twice, or a write followed by two steps with no measurement — the last one only arms, and the message lands before the next write rather than interrupting reading. A measurement has two states: intent is judged from the tool arguments at call time, and the write streak is cleared only when the result comes back, only for that command directory. A failed test still counts, because it yields the line that is wrong. Hard measurements are `lsp_diagnostics`, `verify`, `run_tests` and `run_test`; `ops_bash`, `bash`, `shell` and `pwsh` count only when the command matches a measurement pattern, so `ls` no longer launders a blind edit streak. Signals fire once per episode, re-arm after a real measurement, and are capped per turn; a successful `present` sleeps the gate until the next write, and subagent sessions are skipped. Generated paths (build output, lock files, logs) are excluded. Error targets are read out of real `file:///` ESM stacks, so two different files never collapse into one episode, and `cd <dir> && cmd` counts as a measurement of that directory rather than its parent. Ships the `stuck-gate` skill at bundled rank 600 (a user-level copy at 400 overrides its wording) plus a read-only observation panel at `/better-debug` reporting firings, sessions covered, the first action after each message across six classes, and how many measurement loops actually closed. The log defaults to `$DSH_HOME/logs/better-debug.log` and is created on first write, so the panel has data right after install. Optional host services are subscribed dynamically, so a host without them still loads the gate. It does not estimate tokens saved, and it does not claim the message caused the change. The panel banner counts only the failures after the most recent load (`log.current`), so one old mount failure no longer reads as "currently disabled". Test suites wait for log conditions instead of fixed delays (test-only change in 0.2.5).

Read and edit a Notion workspace from the agent: 10 tools over pages and databases, content handled as Notion-flavored Markdown, and writes guarded by a content-hash version check, a preview diff and one-step undo, so an edit made by hand in Notion is never silently overwritten. Installs with dsh plugin add; no build step. Requires a Notion integration token (or a personal access token) and sharing each target page or database with the integration.

Ship OpenTelemetry GenAI observability and multi-device cloud state sync from one DSH bundle. It reuses and replaces the standalone @loongsuite/dsh-plugin (traces + metrics for sessions, agent loop, LLM and tool lifecycle; do not install both — disable or remove the upstream plugin first), and adds a WebSocket client that pushes full session event logs (header + snapshotEvents) to a relay you run so different local DSH instances share state. On connect it pulls and rebuilds the other devices' sessions automatically (cross-device resume). The sync client itself has zero third-party runtime dependencies; a small Node.js relay is included in the repo. The local /cloud-sync/* HTTP routes (status, sessions, config, resync, delete) have no origin or loopback check, and the bundled relay accepts unauthenticated connections unless AUTH_TOKEN is set — keep the DSH web port off untrusted networks and run the relay with a token.

One-click prompt optimizer next to the composer, with a side-by-side diff panel and per-mode settings. Output length follows your request: ask for a short version and it writes a 350–650 character brief, otherwise it expands to as many as 1400 characters. The prompt templates are original; the official-template comparison mode needs a local Prompt Optimizer installation.

Shows the provider name in the current-model button and its menu, so same-named models from different providers stay distinguishable.