Plugins

Classifies pending tool calls with Jev (TypeSafe AI's non-generative decision model) on tools/pre-execute — confident high-risk calls are denied, ambiguous ones escalated to approval — and auto-approves clearly-granted reversible calls on approval/request behind argument-evidence gating. Also subclasses the session-reference resolver so snapshots keep Jev-scored messages instead of dropping oldest-first. Degrades to stock behavior on any provider failure; shadow mode with a /jev-stats command; TypeSafe or OpenRouter endpoints; 64 tests.

Terminal-native TUI frontend for DeepSeek Harness that presents sessions, models, plugins, jobs, subagents, todos, and provider configuration through Harness capabilities while preserving real terminal scrollback.

Role-based LLM retry & fallback strategies.

Role-based LLM retry & fallback strategies.

Browser automation (Playwright) with a live view, an MCP server exposing DSH agents to any MCP client, and GitHub issue/PR/webhook review tools.

Role-based LLM retry/fallback strategy.

Connects DeepSeek Harness to Multisim MCP for circuit generation, simulation, diagnostics, design optimization, and experiment report export.

Read any page as clean main content: charset auto-detect (GBK/GB2312/UTF-8/Big5), noise stripping, compact text/Markdown, offset continuation, optional SPA rendering, batch reads and site crawling.

StrataGate gives DeepSeek Harness evolving, local-first long-term memory for user preferences, project decisions, conversations, and tool results across sessions; it layers history with user-tunable time decay, reinforces memories through real answer use, shows workspace-aware usage counts, and keeps recalled evidence traceable to its original messages.

Continual self-evolution: versioned, auditable, rollback-safe harness state (prompts, memory, skills, subagent specs) refined from session trajectories, with review gates and hot-reloaded skills.

Minecraft development skills and tools for DeepSeek Harness: 8 skills covering Paper/Spigot plugins, Fabric, Forge and NeoForge mods, addon development against major mods, and Gradle/JDK build setup across Minecraft 1.7.10 through 26.x; the mc_scaffold project generator and the mc_gradle Gradle runner; an intake skill that asks for missing version/platform details before starting; four built-in subagents; and a second Minecraft Architect preset whose mc_codex tool runs the user's own Codex CLI through the codex app-server protocol to draft the skeleton, after which DSH fills the marked method bodies and verifies the build.

Skill marketplace and manager: manage skills from all local coding agents (10+ executors, incl. Claude Code and Codex) in one page and import them into the DSH skill library; built-in market of 6600+ skills, per-skill token-overhead stats, and model visibility control.

Role-based Codex / Claude Code / ACP subagent providers for the DeepSeek Harness — continuable children, durable session recovery, per-role product permissions, and delegation with a permission ceiling.

Durable background child agents on the official subagent seam: start from any session, watch progress in the Web UI sidebar, message and interrupt any time, with per-child tool scoping, persona and delegation-depth caps.

Visually configure an agent's prompt, tools, visible skills, subagents and fallback models, and assemble several of them into a team. It never writes or edits any host preset file.

ClearAI: The Epistemic Loop, native to DSH.

Semantic retrieval, resource management, auto-recall (user + agent dual spaces) and session memory for DeepSeek Harness: memsearch/memfind/memread/membrowse/memgrep/memglob over viking:// URIs, memadd/memremove resource management, memqueue observer status, memcommit session capture, and memlearn skill playbook minting.

Complete port of obra/superpowers for DeepSeek Harness: 14 methodology skills via a native SkillProvider (rank 550), covering brainstorming, planning, TDD, debugging, parallel agents and more, with Chinese-localized docs and zero-build install.

Matt Pocock full promoted skill set (25 SKILL.md: grilling, writing-for-agents, wait-what, TDD, code review, wayfinder, ask-matt router) ported to DSH.

Barrage-meme skill distilled from 22,771 real messages in the 6657 live room: wanjiqi-style banter, CS and DOTA cross-memes, player roasting and casting commentary.

A zero-residence context engine: large tool payloads are evicted from the live context into a queryable store, only pointers stay behind, and any payload can be reconstructed verbatim on demand — so per-turn prompt tokens stop growing with the transcript. It ships four capabilities: zr_recall rebuilds masked originals by tool-call id or session sequence from the persisted session log (the content is not resident, never lost), zr_ledger reports the attention integral and a three-part cost breakdown, zr_compact arms one forced compaction, and zr_fast runs long commands asynchronously and hands back a handle immediately.

The organ kernel of Agent-Body: plugins organized as one body. Each command is turned into a nerve impulse that deterministically innervates the organ which should handle it and names the capability to use; a heartbeat with a variable interval (faster on alerts, slower when idle) pumps operator directives, proprioception and homeostasis warnings system-wide; reflex arcs fire with zero model calls; failures go through deterministic attribution (missing tool / arguments / permission / timeout / network), a prescription, and a re-check loop, with argument errors never auto-retried. A missing organ is compensated by the one with the highest capability overlap, and tool schemas are gated by the current intent (84.7% measured reduction, reproducible in-repo). This entry is the kernel plugin package inside the agent-body repository.

The memory organ of Agent-Body: after 2 minutes of silence it enters light sleep and after 5 minutes deep sleep, where deterministic rules distil real experience into memory cards (pitfalls, playbooks, hotspots, unresolved issues, facts). Incoming commands recall related cards by keyword and inject them into context; cards unused for long decay on a half-life and are archived rather than deleted. It also de-noises body-wide homeostasis warnings: repeats inside a window are counted instead of re-printed, and a warning that keeps recurring without ever healing converges into a known steady-state offset and goes quiet. Consolidation and recall make no model calls.

Imports Claude Code setup into DeepSeek Harness from a native settings screen or CLI, with dry-run previews for skills, commands, agents, hooks, permission rules, and MCP config. Codex and OpenCode remain supported as secondary origins, with collision-safe apply and post-move checks.