Chat with, monitor, and approve your DSH (DeepSeek Harness) agents from WeChat over the clawbot iLink gateway: two-way text/images/voice/files/video, native vision or OCR, context-rotation policies, reminders, and a standalone admin console.
Meta-Forge Workshop: turn recurring work into governed AI workflow assets. MCP server + Claude Code plugin + dsh plugin — staged writes, evals-gated forging, inference engine. 元铸工坊 SOSOFAST：把最佳实践铸成可治理的 AI 工作流资产
Enforce strict checks as dsh harness policy: repeated-failure intervention, a glob-protected critical-path gate that blocks with diagnostics, and an automatic post-write check.
Floating todo progress ball for DSH: live done/total + active task, status colours, draggable, Shadow-DOM isolated, dual-channel sync, history archive (cleanup/restore/copy), pinned sessions, layout switch, discipline injection + first-call gate, 128 shared skins, web + desktop. 常驻 todo 进度悬浮球：实时进度/状态色/历史归档/固定会话/布局切换/纪律注入+首调用硬门/128 皮肤统一，双端可用。
DeepSeek Harness plugin: transcodes images to a base64-safe format (PNG) before a vision request, so gateways like LM Studio (which only accept PNG/JPEG/GIF data URLs) do not reject WebP images.
⚖️ The audit AI coding agents can't skip — deterministic facts (git history, test coverage) and six-gate claim adjudication for code security audits. Not another scanner.
DeepSeek Harness plugin: role-based model routing — planner (root agent) on deepseek-v4-pro, delegated executor subagents on deepseek-v4-flash; ships a prompt section and a pro-flash-routing skill.
DeepSeek Harness bundle: register Claw Orchestrator as an ACP subagent provider — delegate a dsh subagent to a multi-engine council across Claude Code, Codex and Cursor.
System-1 decision models (Jev/Laya/Kev-class) as the governance layer for an LLM agent context lifecycle: growing association graph over session segments, relevance-gated recall with Trace-as-State ordering, and plan pre-ranking — measured in solve rate, cache hit/miss tokens, cost and latency. DSH plugin + harness-agnostic proxy.
