DeepSeek Harness Plugins · Workflow

A minimal infrastructure entrypoint for DSH that probes memory, search, skill catalog, and runtime components, then delivers a PASS/WARN readiness verdict.

A World of Warcraft-style quest system with a quest board, tracking bar, recurring tasks, due reminders, and Agent-driven task creation, progression, and completion.

A DeepSeek Harness skill-gate plugin that presents candidate skills to the user when the system triggers one, defaulting to the system choice if no selection is made within 10 seconds.

Let agents dynamically add, modify, and remove their own tools across restarts with zero new model tools, driven solely by file operations on TS scripts.

Gate an AI agent's completion claim in DeepSeek Harness by running the project test suite first — deny and feed back failures until tests pass.

Structured test-result normalization for DSH, parsing pytest, Vitest, Jest, and Cargo summaries into a stable, uniform result shape.

Adds an /init slash command to DSH that uses a two-phase LLM call to analyze project structure and generate AGENTS.md, with git init, dry-run preview, and thinking-mode toggles.

Integrates FrameFactory public content-production capabilities into DeepSeek Harness as a callable skill for topic breakdown, title generation, script structuring, visual planning, and pre-publish QC checklists.

Provides session-level multi-thread orchestration for DeepSeek Harness, enabling cross-session creation, messaging, and enumeration of ordinary sessions globally.

Adds a start/stop control tab to DSH settings, letting you toggle any Cordis Loader plugin with persisted state and double-click confirmation.

A DAG workflow plugin for DSH Agent that supports creating, testing, publishing, and executing workflows, extracted from Laoboshi Agent Studio.

Separates planning from execution: the main model delegates tasks to an isolated sub-agent, keeping intermediate steps out of the main session to save tokens and enable per-role model selection.