Agent 連携 · DeepSeek Harness プラグイン — 159 ページ

DeepSeek Harness (dsh) 工作区文件浏览器插件：浮层面板浏览当前会话工作区文件树，支持代码语法高亮与 Markdown 渲染预览。

A draggable AI-agent companion pet for the DeepSeek Harness web terminal: reacts to agent states and speaks on approval and completion.

Path registry for DSH (DeepSeek Harness) plugins: inventory all installed plugins/presets/skills with real paths & sources, detect new/removed/upgraded changes, scan C-drive residue references, verify dependency integrity. Agent tools (plugin_registry_list/check/report) + Web GUI panel. Hot-pluggable via profiles cordis.patch.yml.

Native adaptive intelligence layer for DeepSeek Harness: memory, skills, /learn, background review, and curator — no Hermes or Python dependency.

Subagent delegation with per-call model/provider override (host) + multi-panel live streaming subagent viewer (client)

Classic, dependable DeepSeek Harness TUI with session, provider, and model management

Persistent, searchable, per-project memory for the DeepSeek Harness: decisions, rules, and sprint context in a queryable DuckDB file, rules enforced on every model request, with a full management UI in the Web Client

Task management for the DeepSeek Harness: the session todo list, a project task board backed by a queryable SQLite database, and board cards dispatched to the agent as background jobs

DeepSeek Harness bundle for agent-native catalog, contact, deal, receipt, settlement, and human-task workflows

AgentDuel plugin for DeepSeek Harness

AgentFrame Suite — 三件套整合包 + WebUI 控制面板 (记忆/压缩/调度可视化)

Proactive scheduling v2: value network + file/process/http sensors + HMAC webhook triggers + off-peak window

Proactive scheduling for DeepSeek Harness: Aura heartbeat + value network decides when the agent should speak unprompted (official harness is model-driven only)

AgentFrame compaction backend for DeepSeek Harness: semantic + physical KV compression (28.4x) replacing LLM summarization

AgentFrame compaction backend for DeepSeek Harness: semantic + physical KV compression (28.4x) replacing LLM summarization

MemoryDirector plugin for DeepSeek Harness: LLM-driven remember/forget across turns (the official harness has no memory concept)

MemoryDirector plugin for DeepSeek Harness: LLM-driven remember/forget across turns (the official harness has no memory concept)

Governed platform access for DeepSeek Harness — no credential on the machine

Workspace tools for DeepSeek Harness: per-workspace default Agent preset and permission preset for new sessions, using only official extension points (no DSH patches required).

LLM-as-a-Verifier for dsh with the Best-of-N conversation mode built in: rank N candidates with a fine-grained verifier (expected grade over the logprob distribution), on demand via the verify tool or automatically on every turn of a Best-of-N session. Independent implementation of arXiv:2607.05391. By Aispin

dsh plugin: Jev-guided dispatch advice — before a turn starts, a TypeSafe Jev call decides whether the task belongs to a registered agent and which one, then injects a four-line, ignore-if-unsuited recommendation. Jev-optional: no key means silence.

dsh plugin: unattended approval gate — subagent/scheduled/cron approval requests get a bounded answer window, then settle with a fail-closed default instead of hanging forever (per origin or per session)

dsh plugin: cron scheduling — bounded tasks with calendar & interval rules, delivered to live agents

dsh plugin: switch named model profiles (default model, think level, per-subagent models) interactively via the host ask-user flow — one implementation for the TUI and web surfaces, a clean fast-fail on headless