Plugins

Adds a strip above the composer that lists sent-but-unread (steering) text messages and recalls them before the model reads them, one by one or all at once.

Adds /pua to steer the agent to try another approach and verify results before claiming completion, with 15 company styles, persona modes, global and per-conversation settings, and an optional loop that continues until a verification command passes, the iteration limit is reached, or the user cancels.

Reflective self-evolution plugin for local Qwen3.8 27B/3.8 Flash and online APIs: a coding-loop convergence steerer curbs Qwen3.8 27B/Flash repeated-verification deep-thinking to converge faster and cut thinking time; a five-state delivery guard detects and corrects stuck turns and infinite loops; model-agnostic host logic, mostly compatible with online APIs; per-use continual self-improvement with counter-reason self-checks, critic escalation, inverted-index memory retrieval, coding-loop convergence steering, and a five-state delivery guard.

Pauses the agent right before it sends the next model request of a tool round while keeping the composer usable: a bare Enter or the resume button lets it continue, and text typed meanwhile is sent along as a steering message; the pause never interrupts an in-flight API call and the model does not notice it.

A mid-run steer input for DeepSeek Harness Web: while the agent works, an input bar appears above the composer to inject modification requests into the next decision via agent.steer(), without stopping the conversation. Drafts persist per session, unread injected messages can be recalled from the bubble, and an optional guiding prefix is supported.

A simple DSH queue panel for editing, removing, steering, reordering, and bulk removal.

Agentic answer reviewer for DeepSeek Harness: each final assistant turn is re-reviewed by a separate LLM graded 1-100; below-threshold scores steer the agent back with concrete feedback. The score is shown as a chip on the answer's action row (green pass / red needs-work, with retry count and reason on hover). Live gate config from the Reviewer 配置 tab in the conversation view ring above the composer (对话 / 轨迹 / 记忆系统 / Reviewer 配置), from a 127.0.0.1 HTTP server, or from an optional right-sidebar tab via dsh-better-sidebar.

Session bridge for DeepSeek Harness: drive other real DSH sessions straight from your prompt — create top-level sessions in any workspace, message or steer any session, wait for replies and stream them segment-by-segment (chain-of-thought included), resume offline sessions, find sessions across workspaces, run a background watchdog that monitors and schedules a main task (nudge when stalled, correct drift, cancel when stuck), and archive sessions with one command.

Self-contained literature library for DSH with optional Zotero export: full-text PDF fetching (Unpaywall/arXiv/custom URL templates), loose Crossref-backed search, APA/GB-T 7714/MLA/Chicago citation generation, a side-panel PDF reader with highlights and notes, and AI-assisted Q&A/translation steered into the current conversation.

Cross-session task delegation for DeepSeek Harness: spawn durable top-level sessions and send them tasks, read or steer them by base62 session alias, and approve their sandbox escalation requests through an approval bridge that refuses destructive deletes.

Three tools that let any agent, including a team member, list, steer and stop the subagents it started.

Message-level full-text search across DSH sessions via an FTS5 trigram index (user / assistant / steering / tool with per-kind filters), deduplicated results across fork/continued sessions, read-only context preview with auto-scroll to the focused message, one-click jump to the real session, and incremental/full index rebuild with health check; zh/en i18n follows the Host locale service.

Telegram bridge for DeepSeek Harness: bidirectional sessions, conversation steering, home-channel routing, and TTS voice notes sent as Telegram audio messages.

Freshness gate for the agent's test, typecheck, lint and build claims: each verification run becomes a receipt on tools/post-execute, file edits are tracked, and on agent/turn-stopping a claim whose latest receipt is stale, failed or has no readable result steers one more step naming the receipt and the files edited since.

Supervisor-style workflow orchestration over top-level sessions: confirmation-gated batch dispatch, team workstreams, steering running tasks, idle-wait collection and live model-route discovery across 11 task_* tools plus a /tasks slash command.

Remote-control DeepSeek Harness from a HarmonyOS phone: list, switch and create sessions; send prompts (the phone follows the host's busy-enter behaviour, queue or steer); stream assistant text, reasoning and tool steps; answer approval requests and questions; view session images as inline thumbnails with on-demand full-size fetch; switch model and reasoning effort; cancel running tasks; receive push notifications. Traffic is end-to-end encrypted (AES-256-GCM with HKDF-SHA256 key derivation) and travels through a dumb-pipe relay that only sees ciphertext; v1.1.3 adds a session-epoch anti-replay handshake and optional native TLS for the relay. The relay can be hosted on the local machine (managed by the plugin) or on a VPS; the phone app and the plugin ship in one repository. 5 Remote methods.

Resumes an agent turn that ended after narrating its next action without calling a tool: the guard replays the turn log on agent/turn-stopping, asks a judge model one true/false question, and steers the same turn to run one more step.

Separate slash-command group in the Web UI composer: /steer injects text on the agent's next step without interrupting the turn, /new starts a blank session, /skill and /docs wrap /steer with post-turn skill or docs tasks, and custom /steer aliases are configurable in Settings.

DSH plugin: mark messages as invalid so an operator annotation digest steers what a compaction checkpoint absorbs. Cache-preserving: appends after the cached prefix, never rewrites history.

Audit each finished assistant turn with an independent reviewer model and steer the agent to fix failed output, with at most 2 review rounds per turn and SOP-folder standards injected as an extra review dimension.

Adds an /if slash command to DSH: the first line is a condition and the remaining lines are an instruction. A fork subagent that inherits the session tools, system prompt and message history judges the condition and answers YES or NO, and only on YES is the instruction sent to the main agent as an ordinary user message, while a NO sends nothing. The condition, judge prompt and verdict never enter the main transcript. While the agent is busy the command queues with ordinary messages, editable, removable and steerable from the queue panel, and a setting switches to the 0.2.x background mode. The judge runs read-only, and falls back to a one-shot model call over a plain transcript when the host has no context-inheriting fork provider. Supports DSH 0.2.1-rc.5 and later (verified on 0.2.1-rc.5 and 0.2.2-rc.1) and Node.js 20.3 or later.

Stops the agent after N consecutive failures on the same tool. Instead of another retry it pushes a four-part stop instruction: what was tried, the exact error text, the likely cause, and what the user must do. Detection reads real tool results (isError or a non-zero shell exit code) and mirrors them through agent.steer, so the agent sees it as an inbound message rather than a log line.

Deep PTC mode is an open-source agent preset for DeepSeek Harness that keeps the built-in PTC tool surface and replaces the persona with a working contract: a DeepSeek identity and unlimited-token-budget instruction so the model does not skip reads, commands or checks to save tokens; planning by estimated size (internal up to 5 minutes, shown without approval up to 20, approval required beyond); and asking about unclear intent instead of guessing. Waiting is limited in code by wrapping the official shell, job and question tools: foreground commands move to background jobs after 30 s (configurable), a steered user message ends the current wait so the model reads it, foreground pauses over 5 s and long timers in run_code are refused, and questions run in timed mode. The mode derives its preset from the official PTC declaration shipped by the installed Harness and inherits upstream composition changes, so an official PTC update — including removed or renamed modules — does not break existing sessions. A settings card selects the preset executor — Automatic, WSL Bash, Windows PowerShell, native Zsh or native Bash.