Plugins

Detects prompt-injection, jailbreak, and secret-leak patterns on the agent/pre-step, tools/pre-execute, and tools/post-execute seams with allow/ask/block tiers, sanitized defend/detection audit events, a defend_report tool, and a destructive-delete command guard.

Auto mode for DeepSeek Harness (dsh): an Auto entry in the Permissions selector whose approval prompts are answered by OpenGuardrails policy instead of a human — plus the full OGR guard engine underneath. No core changes.

Recoverable file and Git operations for coding agents, with shared recovery and redaction tools, a DSH shell adapter and audit records; optional native text-read protection is off by default.

Per-round thinking-level (reasoning_effort) control for DeepSeek Harness: pick Auto and it schedules low/high/max from the recent tool-call history, or fix a wire level (off/on/minimal/low/medium/high/xhigh/max) manually, with custom wire mapping, a model-aware capability guard, and context-window presets (64K–1M); its per-model capability card pairs with a separate plugin, [dsh-llm-openai-completions](https://github.com/drscrewdriver/dsh-llm-openai-completions) (installable from dsh-market), so custom OpenAI-compatible gateways (vLLM / LM Studio / self-hosted) can try reasoning — the card writes llm-pi-ai capabilities, the adapter drives the wire (compat.thinkingFormat), and this plugin auto-maintains the takeover list.

Update management for DeepSeek Harness and its plugins: semver-aware dual-source (npm/GitHub) checks for the main program and every installed plugin, a locale-aware banner (zh/en) listing updatable items, one-click updates (the main program is backed up, integrity-checked and rolled back on failure; plugins install via a temp directory without touching other packages), and a watchdog-guarded restart after updating. The settings page adds per-plugin version lamps, live update progress, and banner/notification toggles.

Folds degenerate reasoning loops out of the model-visible history so they are not replayed upstream on the next turn. Detects two shapes — one sentence hammered repeatedly, and short-sentence rotation — and replaces the matched span in place; a companion escape valve aborts a turn that will not stop on its own.

Install safety net for DeepSeek Harness: pre-install snapshots, one-click/automatic rollback, guarded boot, and incident reports that auto-trigger agent analysis. 中文: DeepSeek Harness 插件安装安全网（安装前自动快照、一键/自动回退、守护启动、事故报告自动触发 Agent 分析）。

File encoding guard: byte-exact read/write/edit for GBK, Big5, Shift-JIS, EUC-KR, Windows-125x, UTF-16 and UTF-32 files, so editing a legacy file leaves it in its original encoding instead of silently rewriting it as UTF-8; UTF-8 BOM preserved exactly as it was, line endings (CRLF/LF/CR) kept, characters the encoding cannot represent refused rather than written as "?", and non-UTF-8 files listed with candidate encodings so the model can re-read them explicitly.

Remote control: pushes ask / approval / session-end events to an external channel webhook and lets you answer, approve, query status or continue a session from your phone with token auth (npm: `dsh-my-remote`).

Observability and git tooling: a persistent cross-session event audit log with a query API, a resource watchdog sidebar panel, structured Conventional Commits and pre-commit incremental diff review (npm: `dsh-my-observability`).

Skill management: lists skills in global and per-project scopes and toggles each one per project; a disabled skill is not injected into that session (npm: `dsh-my-skill-manager`).

Conversation rendering: every fenced block is rendered by the plugin itself (markdown for text fences, language-banner code blocks with highlight.js for others), mermaid fences become zoomable diagram cards, plus context-body rendering, whole-content copy and a MarkdownView export (npm: `dsh-md-render`).

Memory in global and per-project scopes: memories are injected into the system prompt at session start, a settings panel edits them with confirmation on write, and the agent gets a read-only memory_query tool (npm: `dsh-my-memory`).

Injects a per-session x-opencode-session header into opencode.ai inference requests, fixing the OpenCode Go gateway 400 MissingSessionID error; server-only, no UI (npm: `dsh-my-opencode-session-header`).

Context lens: visualizes token usage of every request (context composition, KV cache hit rate) and enforces per-turn / per-session budgets with warning and blocking (npm: `dsh-my-context`).

Plugin governance: new/updated plugins stage in a candidate area first, then hot-mount after startup — success promotes, failure auto-disables, repeated failures freeze, one-click safe mode, sidebar diagnostics panel (npm: `dsh-my-guardian`).

Browser notifications and a beep when a session turn ends, the agent asks a question or approval is pending; clicking the notification jumps to the session; remote trigger endpoint and SSE channel (npm: `dsh-my-notify`).

Task reliability: auto-retry on model timeouts/failures, auto-continue unfinished tasks, completion verification, thinking-repeat interruption, resume after sleep or restart, autopilot mode and a remote trigger endpoint (npm: `dsh-task-reliability`).

DSH sidebar file-activity tab: records file reads/creates/modifies by the agent and the sidebar (including file ops hidden inside bash commands), per-session isolation, survives restarts, click a file for a floating preview (npm: `dsh-file-activity`).

Renders mermaid/mmd code blocks in the conversation as chart cards with preview/code toggle; the engine is inlined so it works fully offline with no CDN (npm: `dsh-mermaid-render`).

Thinking in Chinese: injects a system prompt so thoughts and replies stay in Chinese, expands the reasoning block by default with interactive collapse, and localizes UI labels (npm: `dsh-think-zh-expand`).

Thinking-loop guard for dsh: observes the llm/stream waterfall and breaks a model that degrades into a loop. Detects two per-call shapes (reasoning-only calls, restated-material calls) and re-fires instead of latching; plus two mid-stream breakers that en

Peak-hour auto session gate: weekend mode + global peak auto-pause + session-level freeze/resume + backend auto-retry, auto-resume on off-peak.

Guarded context compaction for DeepSeek Harness (dsh): the LLM proposes, deterministic guards dispose — eager per-atom shrink (extract/summary/false under verbatim guards) + lazy reference-graph eviction (0-LLM) + byte-exact recall from an append-only log.