Plugins

Lazy senior coding ruleset for DeepSeek Harness: injects minimal-code guidance, adds review, audit, debt, gain and help skills, and switches lite/full/ultra per session. Ships a script to sync rules from a pinned upstream commit.

Campus-network auto-login for DeepSeek Harness: status, one-click login, a local settings page, and an optional autostart watchdog. Includes a Nanjing Audit University preset and a configurable JSON-portal mode.

Track source lineage for web excerpts and structured Agent reports, fold repeated text with recorded provenance, and export auditable evidence through DSH tools.

Left sidebar disk panel for the DSH Web UI: volume free space, per-Workspace folder sizes, moving build artifacts to the macOS Trash with an audit log and restore, sending scan results to an agent, and a host-side low-space watchdog with OS notifications. macOS and Linux only; Trash cleanup works on macOS only, while Linux gets the scan, trend and watchdog without cleanup. Cleanup is limited to a fixed list of build-artifact folder names directly under a registered Workspace.

Translates the English text of DSH approval prompts and ask_user_question cards into Chinese for display only: approval outcomes, option labels, tool names and JSON arguments stay unchanged, and approval audit records keep the original English. The text is sent to Microsoft's Edge translation endpoint (edge.microsoft.com) to be translated, so the content of approval prompts and questions, including command text, leaves your machine.

Security guard for DeepSeek Harness (Cordis): reviews every tool call in pre-execute against SlowMist agent-security rules, denying or flagging risky commands, MCP servers and skill loads, external URLs and repositories, untrusted documents, and EVM or Solana addresses; scans tool results for prompt injection after execution and ships a slowmist_scan audit tool plus an injected security policy prompt. Rules follow the open-source SlowMist Agent Security Skill (MIT).

Observation-only audit for DSH Agent Teams that records per-request cache usage and prompt-prefix rewrites to JSONL, with no tool or prompt changes.

Team presets for AgentTeams: ships audit6 / impl5 / lite2 rosters with seed task DAGs and a per-role report contract, plus a captain doctrine section, so a fresh session opens the right team in one approval instead of a pasted prompt. Needs the engine @nanmicoder/dsh-agent-teams added to the same profile first; the team patch targets that engine's entry by id and replaces its whole agent-teams config.

Local long-term memory core for DeepSeek Harness with plain-text notes, a bounded index, five tools, optional host-model consolidation and auditable evidence. It appends every user and assistant message under 32,000 characters to ~/.companion/cold/<date>.jsonl in plain text, with no built-in cleanup, and adds a short index of its notes to every system prompt; the optional model-based consolidation is off by default and, once on, sends those messages to the session's model.

Forces every active session into the full-access permission preset (this disables the approval gate) and appends approval, sandbox and tool-call events to a local JSONL for after-the-fact review. Audit path is configurable; it never uploads anything.

Replaces identity numbers, phone numbers, bank cards, emails, addresses, customer names and secrets with numbered placeholders before a model request leaves the machine, restores the response, and keeps an encrypted local mapping plus a hash-chained audit log exportable as CSV.

Read-only behavior audit for DSH sessions: flags unsafe actions in 8 categories (A–H) with rules for 14 subclasses (the 41-subclass taxonomy file is not shipped, so a built-in 14-subclass fallback is used), anchors every finding to a real event seq, and gives deep links plus the full tool-call / data-access chain of any assistant reply (parameters, results, files read and written, domains, sandbox and approval context). Every finding is drill-downable.

DeepSeek Harness tools to discover MCP configurations across clients, audit static security rules and probe a named server, using a separately installed mcp-hub Rust binary.

Keyword-rule scan of a project directory for tax-industry compliance gaps ahead of regulatory audits: accounting voucher and ledger retention set under 30 years, invoice-title lines with no taxpayer ID, tax data leaving China with no security assessment or standard contract mentioned, and e-invoice lines with no verification or de-duplication mentioned, with rule sources kept in rules/.

Grades every PowerShell command an agent is about to run before it runs: deletion, overwrite, network download and credential access are four required classes, plus dynamic execution, system and privilege changes. Verdicts are a pure deterministic rule function — no model call, no network, sub-millisecond — with two escalation rules (download+execute, credential+exfiltrate) that jump straight to high. High risk stops the tool call for a second confirmation (in-panel queue or the native approval), deny or timeout means it never runs; medium is logged but never blocked. Every command and its verdict, including the redacted command text, is appended to a local JSONL audit log. Quoted danger strings are treated as data, -WhatIf is respected, -EncodedCommand is decoded and analysed again. Host + client halves, zero runtime dependencies, 55 rules, 136 tests.

Scans a workspace for plaintext API keys, flags files inside git repositories, and reports findings masked so no full secret is ever printed.

Skill and plugin audit console for DeepSeek Harness. It scans local skill folders and installed DSH plugins, checks the packaging rules that commonly break installation, scores each asset, and exports a Markdown fix report.

Capability drift detection: audit a plugin once, store a baseline of what it could do, then after every upgrade report only what is NEW (new install scripts, shell/eval access, outbound domains, credential paths). Deletions and version bumps are not reported, and no score is computed.

Human-in-the-loop self-improvement loop: one objective becomes an auditable work chain that iterates in child sessions while the main agent stays idle.

DeepSeek Harness port of the Impeccable frontend-design skill: 24 design playbooks covering shaping, critique, audit, polish, layout, typography, motion, accessibility and design-system documentation, driven by the reference files shipped in the skill plus PRODUCT.md / DESIGN.md in the project instead of a downloaded engine binary. Registers the impeccable skill; reads project files and writes design artifacts under .impeccable/. The binary-only features (live browser bridge, pixel measurement, font matching) fall back to static variants, screenshots and explicitly declared manual estimates. Ported from pbakaus/impeccable (Apache-2.0).

Windows environment doctor: audits 30 developer tools, installs the missing ones, manages the user PATH, and exports or imports an environment share code so a team keeps identical setups.

Registry of 119 quant and crypto sources covering regulator filings, point-in-time fundamentals, macro series, exchange APIs and on-chain data, each tagged with an authority tier, verification state, five-state point-in-time, revision risk and dataset tags; ships qt_source_find and qt_source_audit tools for source lookup and for citation audits that redact credential-like URL parameters.

Bundled registry of 119 computer-science sources (textbooks, official docs, standards, preprints) across the 17 CS2023 knowledge areas, each tagged with an authority tier, evidence type and refresh interval, plus cs_source_find and cs_source_audit tools that look up a source or audit the URLs in a draft.

Declares a UI layout contract, measures pages in headless Chrome or Godot scenes and declared layouts, and audits overlap, occlusion, overflow, clipped text, undersized targets, dead controls, spacing drift and z-order traps, with a breakpoint matrix and a sidebar panel. Page measurement needs Chrome, Edge or Chromium and the puppeteer-core package installed where the DSH profile can resolve it; Godot scenes and declared layouts need neither.