DSH Plugin | Skills Agents for DeepSeek Harness Plugin
Free Skills & Agents DSH Plugin for DeepSeek Harness: browse, search, download, install skill packs, subagents, orchestration. Verified, traceable sources.
DSH Plugin Hub is a community plugin marketplace for DeepSeek Harness (DSH plugin), built to the official plugin spec and embedded in Settings → Plugin Market. Browse, search and one-click install human-curated community plugins without leaving the app, synced daily with dsh-plugin.org.
dsh-agent-teams turns a DeepSeek Harness session into a captain that assembles durable sub-agent teams, splits goals into dependency-aware tasks, and coordinates via direct messaging—no separate workflow engine needed.
SandBase Harness is a local-first AI agent runtime with sandboxed sessions, MCP tools, memory, credentials, and audit/replay, plus a built-in console, runnable as a DeepSeek Harness (DSH) plugin.
A DSH plugin offering composable task controls for working state, reasoning strategy, quality gates, and self-evolution with human-reviewed cross-session rule learning.
Create persistent multi-model agent teams for DeepSeek Harness with dynamic planning, bounded DAGs, per-member model/tool policies, plus Run Center and Token insights.
Adds an Auto permission mode to DeepSeek Harness, automatically allowing routine operations inside the official sandbox while classifying, asking, or denying out-of-bound risks for balanced efficiency and safety.
DSH Crew is a DeepSeek Harness plugin that lets Claude Code, Codex, and other hosts dispatch work to DSH agents, preserving native subagent progress, tier policy, and a multimodal bridge.
dsh-run2skill is a DSH plugin that automatically turns corrections, constraints, and workflows you explicitly teach the Agent into reviewable, reusable native Skills, saving only after your confirmation.
superpowers-dsh ports 14 core skills (TDD, debugging, planning, collaboration) from obra/superpowers to DeepSeek Harness, registering them in the host skill registry for direct use via the skill tool.
ForkProbe is a DSH plugin for comparing AI skills side-by-side on the same task, generating reports to help you pick the winner.
A DSH plugin that fixes tool-call failures and retry loops caused by incorrect sandbox escalation parameters for third-party models in All Access sessions, zero-configuration.
DSH Agency Agents provides 271 summonable specialist agents for DeepSeek Harness, enabling on-demand expert activation and one-shot subagent summoning.
Agent Team is a DSH plugin for building multi-agent teams with independent models, skills, MCP tools, contexts, and a shared workspace.
A DSH plugin that automatically distills conversations via background subagent reflection, creating or updating skills without manual intervention.
dsh-agent-team brings multi-agent team collaboration to DeepSeek Harness: durable Workspaces, Channels, and Tasks let agents act as true collaborators in sessions.
UltraMath is the first math modeling contest plugin in the DeepSeek Harness ecosystem, driving the full pipeline from problem reading to review with 5 agent presets and a 33-doc model library, featuring triple-blind review and academic integrity gates.
A two-phase DSH plugin preset that anchors the first-turn trajectory with a minimal prompt and few tools, then unlocks the full Standard toolset for high performance.
A DSH plugin adding a subagent monitor panel to the Web sidebar and top-right corner, showing live status, usage details, and an overview dashboard with drill-down navigation.
Vibe Mathematics is a multi-agent math solving and verification framework for DeepSeek Harness, offering four agent presets with resume, human intervention, and cross-validation.
A DSH plugin that registers the godot-4-development skill, enabling AI to handle full-stack Godot 4.x game development within DeepSeek Harness, from planning to export.
How to pick a trustworthy Skills & Agents DSH plugin
A Skills & Agents DSH plugin packages professional capabilities and division of labor — skill directories, prompt templates and subagent orchestration that organize actions into complete workflows, the core of complex task automation. It decides what your agent can do and how it is orchestrated, so look at skill quality, orchestration complexity and compatibility status, and prefer modular, composable plugins.
- Skill quality: read the skill docs, sample outputs and community feedback to confirm it performs well on real tasks, not just demos.
- Orchestration model: how subagents divide work, pass context and support custom flows — this drives reliability on multi-step tasks.
- Compatibility status: skill and agent plugins depend on DSH's agent interfaces — prefer verified plugins whose dshTarget matches to avoid orchestration-protocol issues.
Where Skills & Agents sits in the DSH taxonomy
Skills & Agents is one of the eleven primary DeepSeek Harness categories, focused on capability packaging and division of labor: skill directories, prompt templates, subagents and role orchestration. It is the advanced form of Tools & Capabilities — tool plugins provide single actions, while skills and agents orchestrate actions into complete professional workflows. It is the category that drives DSH toward complex task automation.
「Every part of the product is a plugin — including the model adapter, the tool registry, the session log, and the agent loop itself.」
SourcesOfficial site·Architecture docs·Repository
Related Terms
- Skills & Agents DSH Plugin
- A Skills & Agents DSH Plugin orchestrates DeepSeek Harness agent capabilities — skill packs, sub-agents, role orchestration, multi-agent collaboration and expert prompts.— DeepSeek Harness architecture docs
- Skill pack
- A skill pack bundles prompts, tools and workflows into a reusable unit — one pack is a complete solution for a class of tasks.— DeepSeek Harness architecture docs
- Sub-agent
- A sub-agent is a dedicated agent spawned on demand by the main agent, each with its own role and context, splitting complex work across parallel workers.— DeepSeek Harness architecture docs
- Multi-agent collaboration
- Multi-agent collaboration is a pattern where multiple agents cooperate via messages, tasks or workflows — suited to complex work needing several perspectives.— DeepSeek Harness architecture docs