CATEGORY
Verified manifests and exact versions in this category.
dsh-team-modeDeepSeek Harness plugin: a `team` agent preset whose Lead decides and dispatches while short-lived workers do the writing
dsh-plugin-war-room-protocolWar Room Protocol as a DeepSeek Harness plugin — say the phrase and Sloane boots a 15-seat AI software office.
dsh-plugin-seller-protocolSeller Protocol as a DeepSeek Harness plugin — say the phrase and Mercer boots a 15-seat AI e-commerce office (Amazon FBA + DTC).
@local/dsh-boostQuorum mode bundle for DeepSeek Harness: multi-agent deep reasoning with role-isolated subagents, adversarial verification, persistent return channel, and context budget management.
dsh-max-token-auto-continueCommunity DSH host plugin that continues root-agent sessions after max-token truncation.
@tacrine_f/oh-my-dsh-omo-forkA fork of OMO packaged for DeepSeek Harness: the tri-agent agent preset (Atlas orchestration + Prometheus planning) as a publishable Cordis bundle.
superpowers-preset-dshSuperpowers as a DeepSeek Harness agent preset: the full obra/superpowers engineering methodology in one selectable mode, with its skills and workflow rules scoped to that mode.
dsh-tabbit-brainPrivate text-only Tabbit reasoning conversations inside DSH, with direct gateway requests and background jobs.
dsh-linghun-assemblerSoul’s memory extraction side · Cognitive Loop Team Edition: a five-step closed loop of judgment → retrieval → assembly → evaluation → calibration, supported by role-based sub-agents serving as judge, historian, debater, editor, and clerk. BM25 retrieval + temporal materials + LLM assembly of a “use
dsh-subagent-model-directorSmart subagent model assignment for DeepSeek Harness: define model profiles (intelligence, speed, tags) from the globally allowed subagent model list, then let the delegating session pick the best child model per task — fact-checking/research gets fast low-intelligence models, code analysis gets hig
dsh-plan-mode-plusZCode-style plan mode for DeepSeek Harness: model-initiated enter_plan_mode tool, hard read-only enforcement while planning, and the four-phase planning workflow — builds on the official @deepseek-ai/dsh-plan-mode
deepbotA standalone, always-on agent built on DeepSeek Harness. Slack Socket Mode platform adapter + durable sessions + a memory-ready gateway profile.
@liuyu-f/dsh-plugin-session-cascade-deletePermanently delete DeepSeek Harness sessions from the UI (header button + sidebar row menu) or via the session_delete tool: removes the log directory, projection cache and workspace accounting, cascades into the subagent sessions a session owns (your forked branches are never touched), and stops a r
session-plugin-pickerPick plugins per session from a control in the composer: a clickable entry that mounts the chosen plugins into that session's agent scope
agent-development-constraintsSelf-evolving development constraints for AI coding agents — installs a rule library and injects it into every step.
@shihaoshen/dsh-project-agent-instructionsProject-level AGENTS.md loader for DeepSeek Harness: injects .dsh/AGENTS.md (and custom paths) into the agent system prompt, with a settings page.
dsh-knowledge-workPersistent evidence-backed knowledge workflows and an Agent preset for DSH
dsh-advisor-groupDSH plugin: consult multiple expert advisor models in a retro chat-group card — auto-deepen consultation pipeline with real streaming, SSE replay, and 26 provider presets.
@apherchin/dsh-session-deleteDSH session-row menu "Delete conversation": permanently deletes a session's on-disk data, cascades into subagent sessions, and refuses live sessions (host + client plugin).
dsh-takeoverSession takeover plugin: retrieves sessions from eight external agents (/resume-*), stores the current DSH session in ~/.handoff/pending/ using the handoff: 1 open protocol, and retrieves it at startup to continue working
dsh-linkedin-agentEleven LinkedIn agent skills for DeepSeek Harness (port of Jakeschincariol/linkedin-agent-skill), MIT.
dsh-simple-planDeepSeek Harness plugin that bundles the simple-plan skill: plan, execute with worker subagents, verify every deliverable with memoryless verifier subagents, and fix in at most three cycles. The Skill is shipped by the bundle, so there is nothing to copy into a skills directory.
dsh-plan-guardianDeepSeek Harness plugin that bundles the plan-guardian skill: a mandatory 7-step plan with binary acceptance criteria, execution through worker subagents, and completion validated by independent memoryless verifier subagents that report PASS or FAIL with evidence.
dsh-skill-auto-load-typesafePreload DeepSeek Harness skills selected by TypeSafe AI.
dsh-agent-contractMulti-agent contract workflow engine: contract injection / per-role delegation channels / three output contract tools / three ledger domains / automatic metadata derivation / librarian (on-demand activation) / audit (warning-based).
dsh-reinsDurable Objective orchestration for DeepSeek Harness: Task DAG, durable approvals and waits, crash recovery, and acceptance that runs the tests itself.
dsh-jev-thinkingLet TypeSafe's Jev decide each prompt's thinking level, then apply it to the model request. · 用 TypeSafe Jev 为每个 prompt 自动决定思考档位。
@beibeihk/dsh-agent-observabilityCommunity plugin for event-grounded, privacy-first agent reliability analysis in DeepSeek Harness.
dsh-plugin-android-vscreenLet a DeepSeek Harness agent drive an Android phone on a virtual display — it works in a second screen while your own screen stays untouched. · 让 AI 在安卓手机的「虚拟屏」上干活:手机照常用,互不打扰。
dsh-travel-plannerTravel Planner: A travel planning agent for DeepSeek Harness (DSH)—checks high-speed rail, flights, hotels, and real routes in real time, delivering a single-file HTML itinerary