@omdsh-dev
Side conversation plugin for DSH: /side persistent side sessions (Codex style) and /btw one-off side questions (Claude style) run in temporary forks without writing to main session history; Web UI right-side chain panel with embedded conversation, main session unchanged.
@dongsheng123132
Implements an open task handoff protocol for AI harnesses, enabling verified state transfer between DeepSeek Harness, Claude Code, and Codex via task passports and packs.
@Rxiain
OpenViking retrieval, resource management, automatic recall, and session memory plugin for DeepSeek Harness, with semantic search and memory tools.
@sulfide2085
Unified management of DSH/Codex/Claude AI skills in the DeepSeek Harness settings page: hot toggle enable/disable, one-click discovery and install from GitHub skill marketplace, local ZIP import (dsh-plugin skill hub).
@buhuikongpan
Provides a layered plugin manager for DSH, displaying native plugins by system/WebUI/tool layers and allowing user extensions to be disabled, enabled, uninstalled, and described.
@creght-dev
Codex and agent skills for Creght platform: guides agents on project structure, CLI workflows, page development, CMS/Form/Auth/Func backend, SEO, and site replication.
@Saktawdi
Dynamic Cordis plugin for DeepSeek Harness providing model high-availability failover and multi-agent orchestration with five modes (fanout, pipeline, supervisor, map-reduce, router).
@Classicoke
Plugin suite for DeepSeek Harness that improves execution discipline, with tools for failure interception, reflection, and skill distillation.
@chunsi-w
A closed-loop trajectory policy plane for DeepSeek Harness that manages task episodes, work phases, verification debt, and performance regressions. It includes a policy ledger and adaptive reasoning effort.
@uson1x
Adds an LLM-as-a-verifier to DeepSeek Harness, providing continuous reward signals via select, compare, and track tools. It grades candidate solutions with a model and supports parallel rollout attempts to return the best result.
@SeverusZh
Per-subagent LLM provider/model selection with role templates for DeepSeek Harness, enabling role-based delegation and fallback chains.
@titanwings
Plan review plugin that lets users select exact plan text, add anchored annotations, and return structured feedback to the agent, turning binary approve/reject into precise multi-comment reviews.
