Development & Runtime

Software-engineering method pack for coding agents, with skills for baseline-first planning, systematic debugging, prompt hygiene, verification before completion, and repair/retirement tracking.

Vets third-party plugins before you trust them: static scan for malicious patterns (exfiltration, credential access, obfuscation, persistence) and over-privileged path use, transitive-dependency coverage, official-package hash baseline for supply-chain tamper detection, and an optional plugin-tool call gate.

Prometheus metrics and a ready-to-import Grafana dashboard for DeepSeek Harness. The plugin observes public DSH lifecycle seams, keeps metric series bounded, and avoids exporting conversation or tool payloads.

Loader-independent startup recovery for DSH Web that detects likely broken plugins, temporarily skips them, and restores only Boot Guard-managed changes.

Unifies every DSH extensibility seam behind one declarative DSL: a runtime `fabric` service, `fabric_extend`/`fabric_inspect` tools, and a capability-graph settings page.

The Fabric/Mixin extension layer for DSH as a self-contained workspace of three packages plus an installable profile bundle carrier. The workspace mirrors the upstream fabric split: a pure-Cordis pair (cordis-fabric, cordis-fabric-api) and the DSH integration package (cordis-fabric-dsh) that supplies the Host and browser facades, the package invariant, and the profile bootstrap.

Code Smell Radar for DeepSeek Harness. A read-only DSH plugin that statically scans a repository for common code smells: TODO/FIXME/HACK debt, stub implementations (NotImplemented / 未实现), over-long lines, oversized files, and duplicated code blocks — then outputs severity-sorted, actionable fix suggestions. Install with: dsh plugin --profile web add "github:lucky8197/dsh-code-smell#main".

Native interactive diff review for DeepSeek Harness. The plugin reads a Git unified diff, parses it into files, hunks, and line locations, presents its own terminal review interface, and queues submitted annotations as the Agent's next durable follow-up turn.

A hands-off safety net: directories the agent deletes go to a trash folder first and symlinks are never followed, with zero approvals.

Where the official repeat-tool-reminder is advisory — it nudges an agent that repeats the exact same call — this guard vetoes: after two non-converged experiments on the same problem it injects the three questions, denies non-investigative tool calls until the model answers them in its reply text, and blocks further same-direction shots.

Appends local hash-chained JSONL receipts for tool results and session events without storing prompts, tool arguments, result text, or raw session IDs.

Per-model token cost tracking with configurable cache-hit/miss, output and peak-window pricing, a live session cost bar, and unconfigured-model flags.

A UX walkthrough plugin for DeepSeek Harness (DSH): AI simulates target users to uncover UX problems during development—before they reach production—and provides concrete optimization suggestions. Scope: React + TypeScript / React + JavaScript / Vue 3, CSS/layout analysis, and optional browser evidence when the current Harness session can open the application.

Pin Git commits to the environment's own author identity; env-var injection overrides all `git config` settings.

An all-mode tool failure recorder for DeepSeek Harness: whether the agent runs in native mode or PTC (Code Mode), any tool failure is automatically written into the machine-maintained section of a skill — normalized-dedup, counted, deterministically ranked, TTL-pruned, and redacted — so the next session's model sees the most common failure causes when it loads the skill. Fail less over time.

A self-contained standalone repository template for an ESM Cordis plugin. Every source file, compiler setting, test fixture, contributor instruction, skill, and build helper used by the repository is inside this directory; every development input resolves below this repository root.

Read-only health clinic for the installed DSH plugin set: loader health, dependency integrity, version compatibility, install-script risk, duplicates and patch integrity, delivered as a model tool, a Settings dashboard and JSON reports.

Community distribution: TUI, desktop, and Web UI as one bundle with layered installation.

Git status, diff, log, commit, branch, and optional restore-point tools for DeepSeek Harness.

Remote Host Tunnel Manager: automates the "local browser → dsh web on a remote Linux server" link — remote port allocation with a registry, systemd supervision, resilient SSH tunneling, local URL output, and full lifecycle management. Built for single users and for teams sharing one server.

The nono (Landlock/Seatbelt) backend as an installable DSH profile bundle. This standalone package contains the sandbox provider, its invariant companion, the bundle patch, and the vendored @dsh-external/nono-ts native executor carrier.

dsh-agent-budget gives one live agent session, or its complete local descendant tree, a durable Token limit and absolute deadline. It reserves capacity before every attributed llm/stream provider attempt and replaces that estimate with provider-reported usage after the stream settles, so concurrent child agents cannot all spend the same remaining balance.

LAN access for the Web GUI: 0.0.0.0 bind plus a crypto.randomUUID polyfill for non-secure (LAN HTTP) contexts.

Runtime dangerous-operation policy, canonical output redaction, and security-review workflow for DeepSeek Harness.

Windows desktop shortcut for dsh web: launches the server, auto-opens the browser when ready, and stops the service when the Harness window closes.

dsh-tmuxctl — the control plane for tmux: list, drive, capture, split, swap, run, and watch the panes you already have open, with safety by default (DeepSeek Harness plugin bundle)

WebUI authentication enforced at the HTTP/transport layer: four-layer login gate (resources, plugin bundles, /api, WebSocket), server-side sessions with HttpOnly cookies.

LAN / remote-access control, frpc NAT tunneling, a panel password gate, and mobile UI adaptation for the DSH Web GUI.

The DSH loader tree is composed at boot, so adding a static client plugin (a UI panel, a widget, a dashboard) normally requires restarting dsh web. This plugin fixes that: it watches a hot directory and installs/updates client plugin bundles at runtime on every open page.

Update checker plugin for the DeepSeek Harness (dsh) Web GUI: shows the current version in the session top bar, auto-checks the latest version on npm and prompts you to upgrade when a new one is available. DeepSeek Harness（dsh）Web GUI 的更新检查插件：在会话顶栏显示当前版本号，自动检查 npm 上的最新版本，有新版时提示升级。

Read-only runtime management panel for the official DSH MCP client: connection status, registered tools, errors, and reconnect counts through the /mcp command and a Settings tab, with sanitized display and enable/disable patch suggestions.

GUI in the DSH settings panel: toggle/delete MCP servers, browse and trash skills, and list built-in plugin packages — hot-applied without restart.

Git Hygiene Checker for DeepSeek Harness. A read-only DSH plugin that inspects a repository's hygiene: merged-but-not-deleted branches, stale branches with no commits for months, oversized tracked files, untracked files, and uncommitted work — then outputs a health report with cleanup suggestions. It never deletes or modifies anything automatically. Install with: dsh plugin --profile web add "github:lucky8197/dsh-git-hygiene#main".

dsh-trace stores DeepSeek Harness session telemetry in a local embedded yiTrace database. It observes records after the host's telemetry/record waterfall, projects each DSH turn into one yiTrace trace, and writes SDK-native start, log, and end events through yiTrace's Node-API database. No HTTP server, port, or token is required. The plugin is opt-in, adds no model-visible context, and lives outside the DeepSeek Harness monorepo.

Redact supported credential patterns from outbound telemetry copies before configured backends receive them—without rewriting canonical session logs.

A Claude-Code-auto-mode-like permission classifier for DeepSeek Harness. Adds a fourth permission preset auto (Autonomous) beside read-only / workspace-write / danger-full-access: tool calls are classified automatically — dangerous operations are blocked before they run, safe ones flow, and sandbox escalations are decided by the classifier without a human watching the approval prompts.

This repository provides Codex/agent skills for Creght. They help agents understand Creght project structure, CLI workflows, page development conventions, CMS/Form/Auth/Func backend/SEO features, and other platform capabilities.

Runtime-switchable model output styles with Claude Code outputStyles parity: a /style command, per-session persistence over the output_style domain, systemPrompt injection, six built-in styles, custom Markdown/JSON libraries, and a web picker.

Code2Skill is a collection of installable Agent Skills that helps coding agents understand business functionality in user-authorized frontend, backend, or full-stack source code, then generate Functions, MCP tools, workflow Skills, and offline tests for other agents to use.

In-app plugin store for the DSH Web UI: browse, search, one-click install, compat badges.

- Full token coverage — 181 tokens with zero duplicates: the static color ladders, every alias, the specific tokens and the shiki syntax colors are all themed, so no default DeepSeek blue-gray leaks through. - Classic mauve brand — the brand color follows the Catppuccin way: mauve, not the built-in blue. - Your choice is remembered — the selected flavor persists per browser in localStorage and is re-applied at startup, even when the host re-asserts its own preference. - Component-level accents —

Compare multiple skills on the same task and pick the winner.

MCP Lens lets DeepSeek Harness search and call 1,000 remote tools through two stable model-facing interfaces. Instead of sending every tool schema on every turn, it reveals exact schemas only for a small ranked set when a tool is actually needed.

Turn-scoped “Allow for this task” approvals: automatically allow matching `danger-full-access` escalations only for the current task, then expire.

Login gate for the DeepSeek Harness (dsh) web surface: password or shared-token authentication, session cookies, rate limiting, and a user-management CLI. | DeepSeek Harness (dsh) 网页版登录门插件:账号口令或共享令牌认证、会话 cookie、登录限速,附用户管理 CLI。

A permanent Cordis plugin for the DeepSeek Harness Web GUI that auto-checks for new DeepSeek Harness releases AND installed third-party plugin updates (the former standalone dsh-plugin-checker was merged in v1.1.0), asks the user, and one-click updates with success/failure feedback.

A TPS metrics plugin.

Use a DSH profile that already contains at least one third-party bundle. Replace web with your profile name. The commands below are split between two terminals because DSH normally stays running:

Model-driven /init command: the agent analyzes the current repo and generates/updates a high-signal AGENTS.md, with a persistent system-prompt reminder to keep it maintained.

The community keeps reporting that DeepSeek models do their best work in DSH's Minimal preset. On Windows that preset does not run at all. Its persistent bash needs a PTY, and the stock subprocess runtime resolves a platform process inspector that throws on win32 before node-pty is even reached. Every Windows user has been locked out of the mode the model is best aligned with.

Toggle or delete plugins from the Web settings panel, with runtime enable and disable persisted to the profile.

DSH self-updater in the settings page: one-click check/pull (`git pull --ff-only`), auto background checks, version diff and changelog preview with a red-dot reminder.

Skill for packaging DeepSeek Harness into a Tauri 2 desktop app (macOS + Windows): dual-platform install playbook, China mirror bootstrap with a subagent timeout sentinel, admin-free Windows toolchain recipe (no VS Build Tools), and three-path acceptance checklists.

DSH 上下文注入审计插件：统计 AGENTS.md 指令链/技能目录/工具 schema 的 token 成本，检测重复与冲突；Web UI 圆环面板 + context_audit 工具。Context Doctor for DeepSeek Harness: audit instruction-chain / skill catalog / tool schemas token cost.

- id: tool-approval name: dsh-tool-approval config: include: [fs, web] exclude: [taskoutput] reason: tool execution requires your approval

DSH Testkit packs your plugin, installs it beside an exact DSH version in a disposable environment, boots the real host, exercises deterministic capabilities, uninstalls the plugin, reboots the same profile, and retains reviewable evidence. No model call or model API key is required.

@loongsuite/dsh-plugin is a standalone, open-source observability plugin for DeepSeek Harness (dsh). It observes DSH's native session, agent loop, LLM stream, and tool lifecycle, converts them into OpenTelemetry GenAI traces and metrics, and exports standard OTLP/HTTP protobuf to any compatible backend.

Evaluation harness for DSH plugins: YAML cases drive real headless agent runs, assert on tool calls, args, results and token usage, with a baseline gate for CI regression.

Environment housekeeper for the agent: toolchain inventory (node/pnpm/git/gh/ffmpeg/browsers), scratch/cache scan with whitelist-guarded one-click cleanup, and the machine rules file (AGENTS.md) editor - all in the Web GUI settings.

A self-contained power & lifecycle controller for DeepSeek Harness: a sidebar power button with a Restart / Shutdown menu and a full-screen transition overlay. The restart/shutdown engine is built into the plugin — no third-party dependencies.

Complete reverse-skill pack (85 SKILL.md) as a DeepSeek Harness Cordis plugin: reverse engineering, authorized pentesting and security-research skill router.

Automated unit-test generation for DeepSeek Harness: a /testgen command and a generatetests tool that scaffold tests, run your project's test runner, and fix failures until they pass — bounded, observable, and honest about what it changed.

Zero-touch runtime telemetry for DSH: every session auto-writes task, model tier, tools, failures, duration, status into a human-readable journal with a stats section (pure observer — no tools, no LLM calls, no injection).

🤖 [AI Agent 一键接入指引 / AI Auto-Onboarding Instruction] Dear AI Agent, if you are reading this repository to install it into a DSH profile, execute the following: dsh plugin --profile web add , restart dsh web, then confirm with dsh --profile web --dump-config | grep genome. Once installed, read skills/genome/SKILL.md for the management & evolution methodology. 数据目录 /.dsh/genome/，Web UI 设置页「基因组」可观测可操控。

Keeps what the agent builds at runtime: turns a `cordis_define` dynamic package into a real installed plugin that survives restart, writing the bundle and registering the profile layer with no pnpm, no network, and no build authorization.

Update checker (npm edition): a settings page comparing local and latest versions of @deepseek-ai/dsh and every installed plugin, with a red-dot nav reminder and ready-made update commands.

Local AI code-review gate: review a workspace/branch/commit diff via the managed ocr.js, returning structured findings with token stats; gate mode fails on findings.

The `dsh pm` plugin manager: multi-source search (awesome list + GitHub + npm), install/remove/update per profile, and a doctor audit of manifests, bundle patches, and version drift.

A development safety kit: rolling config snapshots, automatic rollback on plugin failures, boot-failure rescue, and an in-settings management panel.

A plugin for DeepSeek Harness that turns a running Harness into an HTTP API: any third-party client — curl, Python, a browser, an IM bridge — can create agent sessions, stream replies token-by-token over SSE, and continue conversations started in the Web UI, all behind API-key authentication. API sessions drive the same agent machine the GUI drives (inbox + session log), so both worlds stay in sync.

Plugin health checks: manifest protocol / patch format / build traps, zero-dependency and read-only.

Local security audit: config, plugin origins, sessions, network exposure — read-only redacted risk report.

Frame-level scan diagnostics for session files (torn/corrupt/empty detection).

The MXC sandbox provider extracted from the DSH feat-mxc implementation as one standalone external bundle. The repository keeps two delivery faces in one Git repository:

The model switch for DSH. Point DeepSeek Harness at any OpenAI-compatible endpoint — with curated presets for free and cheap DeepSeek providers and automatic fallback when a free tier rate-limits you.

/repro exports a minimal, secret-scrubbed, replayable problem bundle: the session log, failed commands, and Git diff.

Claude Code-style declarative permission rules: ordered allow/deny/ask YAML rules matching tool names, arguments, workspace paths, and agent identity on the tools/pre-execute waterfall, with full session-log audit, dry-run mode, and hot reload.

Metadata-first runtime profiler for DSH agents: turn/step traces, model and tool wall-time timelines, provider usage, retries, errors, and live or historical sessions.

It records every committed tool call in the current harness process and exposes a toolstats tool, so the model (or you) can ask at any point:

Ecosystem infrastructure: a thin browser console for managing official repository plugins (zero patches) plus a make-dsh-plugin skill for guided plugin development.

Private, out-of-tree runtime bridge between Multica and the public DeepSeek Harness. It exposes a versioned JSONL protocol over stdio and composes over @deepseek-ai/dsh-base. It does not require changes to DeepSeek Harness.

Restart DSH: configurable restart method (Node native / legacy PowerShell), post-restart continue prompt, optional watchdog auto-relaunch.

Offline diagnostic for DSH: 19 checks across env, profile, and session state, with a settings "Doctor" panel and a read-only JSON API.

Particle-network background overlay for the DSH Web shell: full-screen, click-through, zero runtime dependencies.

Catppuccin themes for the DSH Web GUI: Latte, Frappé, Macchiato and Mocha registered into the native theme runtime, one-click switching with a persisted choice.

Weekly Digest Generator for DeepSeek Harness. Aggregates the last N days (default 7) of git commits, DSH session activity and daily memory entries into a Markdown weekly report — commits per day, top contributors, recent commit list, per-project session counts, and daily memory highlights. Read-only end to end. Install with: dsh plugin --profile web add "github:lucky8197/dsh-weekly-digest#main".

Collapsible right-docked web terminal (xterm.js) bridged to a host PTY over WebSocket: bash/zsh switching, resizable panel, and fold/unfold that keeps the session alive.

Contract regression harness for plugin authors: a scripted LLM adapter drives the real pipeline in isolated child processes with session-event assertions (load/contract/behavior/fault layers); zero API key, CI-ready.

Verification-driven self-evolution loop: failure logs become verified AGENTS.md rules; in-session plugin (evolve_learn / evolve_apply / evolve_touch / evolve_recall), tool-verify gate, rule lifecycle, local recall.

Modifies DSH plugin code at runtime, with ordered patches, inspection, and hot reload.

Fine grained per tool permission rules for DSH at the tools/pre-execute gate, deny and ask lists in Claude Code rule syntax (Bash(rm -rf:*), Read(_secrets_), mcp__server__tool), works standalone without migrating.

Reverse tunnel for the DSH Web GUI: exposes a `dsh --profile web` deployment at a public frp URL and switches the directory picker to the in-app browser, so a phone or remote machine can open and manage the workspace.

Self-editing desktop workbench for DeepSeek Harness: a full dsh web replica with an edit mode in which the agent modifies the workbench (patch layers, plugins, UI) with hot reload.

One-command bundle of the three AgentFrame plugins: memory-director, compaction, and aura-scheduler together.

DeepSeek Harness environment diagnostic tool: dsh_doctor checks env, profile, config, bundles, mount, port, health, and disk

Close and restart controls for DeepSeek Harness Desktop and Web on Windows, in Settings → General.

Move a whole Claude Code setup into DSH. Skills, MCP servers, hooks, subagents and permission rules, with a dry-run moving estimate and a migration diff report; installs as a plugin exposing a movein_from_claude_code tool so the agent can do the move for you.

Self-healing runtime layer: a resource control loop with strategy switching and repair escalation, error-storm and fiber-failure insight, cross-session knowledge carry-over, a self-healing dashboard and 7 tools.

Layered plugin manager for DSH web: native plugins grouped read-only by system/WebUI/tools, user extensions with enable/disable, register, uninstall and editable descriptions.

Swaps the filesystem and bash providers for a mirage virtual workspace: file tools and shell commands run over mounted resources (RAM, S3, Redis, Slack, Gmail, Notion, Postgres) instead of the host disk, with per-mount read/write/exec modes, per-command sandbox routing (monty, pyodide, quickjs in process; docker, e2b, daytona remote), and installed CLIs (git, gh, slack, linear, ntn, gws, or one you register) as head words in the virtual terminal.

Update copilot for DeepSeek Harness: tracks the dsh core, bundles, and every profile plugin (npm + git dual channel); changelog briefs for agent-assisted decisions; confirmed updates only · DSH 更新助手：追踪 dsh 本体、bundle 与全部 profile 插件（npm+git 双通道）；变更简报辅助升级决策；仅执行确认过的更新

Writes local JSONL summaries of per-turn tool counts and coarse verification signals without storing prompts, tool arguments, or result text.

自进化插件：agent 在 session 内随对话给自己长出/剪掉能力 —— evolve_add 热挂载持久化 cordis 插件（下一 step 工具即可见），evolve_remove 可逆卸载，重启自动恢复

A DSH profile bundle for the fail-closed microsandbox microVM capability. The root package contains the ctx.microsandbox provider and exports the model-facing tools as @deepseek-ai/dsh-sandbox-microsandbox/tool.

Visual browser feedback for DeepSeek Harness. /annotate asks the companion Chrome extension to enter selection mode; each selected element contributes a selector, DOM facts, computed style highlights, accessibility data, a comment, and an optional viewport screenshot to the agent's next turn.

Reusable Markdown prompt profiles for DeepSeek Harness. A profile can select a provider, model, and reasoning effort for one turn, render command arguments into its body, and restore the previous selection afterward.

Evidence-backed inspector for DSH Web built-in capabilities: runtime/config provenance, compatibility and drift diagnostics, plus fail-closed controls for nine reviewed UI leaves.

This is a dual-face theme plugin for DeepSeek Harness (dsh). It registers four faithful palettes into the built-in theme runtime, so they appear as selectable skins in Settings → General → Solarized / Selenized themes.

Secure remote access for the official Web UI: authenticated LAN/tunnel gateway with QR + HMAC one-time pairing, per-device sessions and revocation, mDNS discovery, and a fork_session tool.

Static pre-install security auditor for plugin bundles: lifecycle scripts, dynamic execution, credential-exfiltration combos, and patch-layer hazards, with zero dependencies and in-memory tar parsing.

Document-Code Consistency Guard for DeepSeek Harness. A read-only DSH plugin that audits your versioned Markdown design docs against the actual repository: header version numbers, changelog tables, directory-structure trees, module/test counts, and cross-document references. It detects drift (docs saying one thing, code being another) and outputs actionable, severity-sorted fix suggestions. Install with: dsh plugin --profile web add "github: /dsh-doc-guard#main".

Automatic type-check diagnostics: after the model edits or creates TS files, runs `tsc --noEmit` in the background (debounced per project) and reports errors via the `code_check` tool — grouped by file with line/column, severity and TS error codes, closing the "edit → check → fix" loop.

The Token Anxiety widget tracks the cost of every task in your conversation in real time. It runs in the band under the chat composer and shows DeepSeek pricing status (peak/valley), per-task token usage and cost, and a one-click analysis of where tokens were spent. It is installed as a bundle with no additional dependencies.

Login gateway for the DSH web UI: password door with first-run setup, bcrypt + at-rest encryption (AES-256-GCM/HMAC), brute-force lockout, audit log, TLS 1.2+ with 80→443 redirect, CSRF, anti-framing.

Plugin compiler with a blueprint registry (scaffold → validate → deploy) exposed as a runtime service, three model tools, and a blueprint-gallery settings page.