Plugins

DeepSeek Harness plugin that stops shell commands that would hang a turn (editors, REPLs, pagers, watch modes, dev servers, tail -f, interactive logins, git commit without -m) and tells the model the non-interactive alternative.

DeepSeek Harness plugin that blocks file edits which replace real code with placeholder comments like '// ... existing code ...', and whole-file rewrites that silently drop most of a file.

Keeps the shared default model pinned, so switching the model inside a chat no longer rewrites the global default and no longer lowers that default's reasoning effort.

ChatGPT-desktop-style model picker whose thinking-level slider carries a particle trail, plus a visual-only fast-mode toggle and a hold-to-lock for it.

Aborts a DSH reasoning stream that starts repeating itself instead of letting the turn stall: five rules over the streamed reasoning text (periodic unit, repeated block, repeated line above a share floor, 64-gram density, and a long trailing run of punctuation or whitespace), plus a JSONL fire journal, a Web GUI log panel, and opt-in recovery that retries with a corrective message.

Makes the built-in web_fetch tool work when Clash / mihomo runs in TUN mode with fake-ip: it lets the fake-ip address ranges (198.18.0.0/15, fdfe:dcba:9876::/64, 2001:2::/48) past the public-address check that otherwise rejects them, while loopback, private and link-local addresses stay blocked as before. Works with no proxy environment variables and no Clash changes.

AI companion with a learned body clock, real offline sleep, its own memory, and self-reported affect delivered through the same judge math as the pi version.

Project-level memory built around a ledger whose structure the AI designs per project; the plugin only handles detection, injection, write-back, archiving and git. Memories are split by meaning: the active ledger holds current state, the turn journal is append-only, and decisions, the goal tree and lessons are one-card-per-file (with a multi-dimension index over the card folders); the fundamental principles file can be changed only by the user. There is a single compaction mode: before compacting, the AI sorts the work into three buckets - what must stay in context and is re-injected verbatim, what is written to the project's _stash/ for later lookup, and what is dropped. When a conversation is opened in a container workspace holding several projects, it can be locked to one sub-project (the lock lives in the plugin's own home and survives restarts; it leaves no marker in your project). Also an activity log with five levels, where L4 stores the user's own words verbatim (never injected, queryable on demand). Before each compaction a verbatim capsule (user quotes, raw errors, changed files) is written to disk, and both the capsule and its source material survive a restart; a handoff document lets a new conversation resume the previous one. 21 tools, 12 subcommands, 39 settings (one of them switches whether the settings panel or each project's manifest wins). Runs locally, no runtime dependencies.

A chip at the left end of the composer tool row showing whether DeepSeek bills at peak or off-peak right now. Clicking it opens a card with the current state and rate, the next switch with a countdown, and why this instant is in that state: a weekend, a named statutory holiday, a 调休 make-up workday, or simply outside the peak windows. Times are resolved in Beijing time with an explicit UTC offset, so the verdict does not depend on the machine clock; a hard-coded 2026 Chinese holiday table carries the 33 holiday days and the six 调休 workdays, and a year beyond the table says so instead of guessing.

Manage skills and MCP servers from one DSH settings panel. Skills are read from disk: browse, search, preview the body, create, rename and delete, with a backup written before every removal. Global MCP servers are the real cordis.patch.yml loader rows; per-project entries mount only for sessions in that project. Each expanded card lists the tools that server actually registered, and a per-server allow/deny rule keeps only the ones you need — a blocked tool never enters the tool registry, so it costs no context on every turn. Every MCP card carries a connection status tag (connected with a tool count, retrying, gave up, activation failed) plus a one-click reconnect, so a server that fails at startup is visible instead of silently missing its tools. Imports existing configs from 26 other MCP clients.

Local TXT reader in a right-sidebar tab plus a spoiler-safe AI reading companion: the model only sees what you have already read — the current chapter, the tail of the previous chapter, and a per-book background file it grows as you go. Excerpts, thoughts and the companion replies become structured Markdown notes exportable to Obsidian; raw text unlocks only for a book you mark as finished.

Replaces the reasoning-effort dropdown with a draggable snapping slider, and draws 13 canvas effects across five style families (thrust, fluid, fantasy, tech, texture) with three tiers that unlock as the effort level rises. Custom mode combines 10 particle forms with 8 flow trajectories; 9 accent presets and a hex picker recolor every effect, and optional key sounds cover two voice clips, a synthesized duck squeak, a mechanical click and a local audio file.

Element annotation and review panel for the DeepSeek Harness web client: pick UI elements in a live app preview, write notes, and send them to the conversation as one structured block

Builds, tests, archives and runs an Xcode project or workspace on iOS or macOS through xcodebuild — a live filterable build log, scheme and destination pickers read from the project, and log streaming for simulators and USB devices, including iOS 16 and earlier over libimobiledevice, whose app logs stay readable while the device is locked. macOS only.

English edition of Weiwen's Law (KISS's Law): the first and only domain-agnostic white-box causal adjudication middleware for AI agents. Unlike domain-specific causal-inference tooling (statistics/economics/ML) or evidence-gated agent firewalls, it adjudicates *every* tool call along a deterministic structural causal chain (R->S->D->H->M) before execution - across any domain, with no training or per-domain tuning. It blocks destructive and credential-file operations, escalates repeated boundary violations, cuts faulting links until verified-fixed, and gives risk verdicts on un-auditable execution; runs fully local with zero API cost.

Blocks an agent from reading, searching, writing, or editing chosen files and directories, keeping their contents out of the conversation — including sensitive-word files whose text can make later provider requests fail with a 400. Rules are picked from a file/directory browser on the plugin page.

Shows a Chinese translation under each English thinking block in the DSH web chat, without modifying the session transcript; per-chunk caching, a configurable translation model, and a settings panel.

Turns the MacBook keyboard Caps Lock light into an agent status light: on while the agent works, flashing when it waits for your approval or answer, off when idle.

Mind-map project planning with blocking drift protection: plans, modules and tasks as a dependency DAG with topological execution order, an agent position marker, a human approval gate on every board edit, git cross-checks, and evidence-gated task completion.

Per-project memory for DeepSeek Harness. Three Markdown files per workspace — PROJECT.md, DECISIONS.md and SESSIONS.md — are auto-created on the first step of any session in that workspace, re-read from disk and re-injected only when they change (and re-injected if compaction dropped them), and written back by classification through a memory_checkpoint tool, with a bounded end-of-turn nudge when a turn did work but recorded nothing. memory_read and memory_search read the live log and the archive, the latter over a zero-dependency BM25 index. The session log is capped at 200 entries with the overflow moved to SESSIONS-archive.md — never injected, but visible, readable and searchable, and restorable one entry at a time. Settings page: a workspace list with a path filter, read-only tabs for the three memory files plus the archive, reload, export/import of one workspace's memory as a JSON bundle, a two-step clear, a panel that rewrites or removes the AGENTS.md instruction block, and a PROJECT.md staleness reminder. A composer status chip shows that workspace's sync state and creates the files on click. Zero dependencies — no vector store, no embeddings, and no model calls.

Feeds every dsh prompt into the local agent-feedback-loop feedback-memory pipeline — captures dissatisfaction, reviews it offline into reflections, and attaches the compiled recurrence-rule context back to the turn, never blocking a prompt. Requires the agent-feedback-loop runtime; running agent-feedback-loop install afterwards wires existing dsh profiles automatically.

Peak/off-peak billing clock with a provider-neutral account readout (DeepSeek balance, subscription plan windows, relay credits), per-turn spend measured from the provider's own counter, and a /hist token-history scene split per provider for dsh-tui

Project-based learning coach: a subject-agnostic teaching engine plus swappable domain packs (Unity/C# included), shipped both as a DSH skill and as an installable bundle. The skill protocol enforces attempt-first teaching and graded hints; a zero-dependency validator rejects ability claims their evidence does not support and blocks stage acceptance while blockers are open.

Global prompt-injection / context-virus defense for DeepSeek Harness: scans tool arguments, tool results, pre-model messages and the outbound stream; quarantine/block/monitor modes, canary trap, and an evolvable on-disk signature library with learn/import/export.