Plugins

Engineering workflow agent preset for dsh: five gated phases (requirements clarification, plan approval, TDD, parallel subagents, verified finishing) with six workflow skills adapted from obra/superpowers.

Cross-session plaintext memory: deterministic BM25 keyword recall (no embeddings), memory_save/list/search/confirm/forget tools behind a human-confirm gate, and global + project-scoped JSON stores that follow your git repo.

Cross-session archive: overview workbench, folded timeline, subagent lineage, full-text search, rule+LLM summaries, ZIP export, live tracking. Native session-page embeds: turn rail, docked right panel (live mode, pinned to original session on switch), rename (official-synced), conclusion digest, health monitoring + persistent header chip (idle / active / tool-wait / stalled). dsh v0.1.2+: fuses with the official TurnNavigatorRail — vector-icon info chips (duration, tool calls, errors, thinking) plus lazy tool distribution inside the official preview card, right-click opens the workbench; the classic piano-key strip is on by default and delegates unloaded-turn jumps to the official auto-load engine; the official rail can be hidden entirely. Settings dialog section with 10 tunable behavior params.

Deterministic keyword routing, allowlisted model switching, and an isolated image_gen channel for DeepSeek Harness.

Parse/query/aggregate/convert CSV (RFC 4180) with a zero-dependency state-machine parser.

Circuit breaker for idle agent turns: watches agent/assistant-stream and cancels a turn that produces only reasoning — no text and no tool call — after 45s, after 80000 reasoning characters, or on a degenerate repetition pattern (exact repeat unit, sentence-template repeat, sparse repeat, or 6-gram density). On trip it injects a continue instruction that requires the next action to be a tool call, and ships 11 regression cases plus per-gate tuning.

Behavior-drift reports for DSH sessions: compares two session logs and reports tool-sequence alignment, tool mix, reasoning volume, timing, retries and file targets, with an optional CI threshold gate. Zero dependencies.

Persona engine for DeepSeek Harness: one config.json drives the AI's self-name (per model, per tier), how it addresses you, relationship stance, character text, per-item work contracts with individual toggles, thinking-chain language, appearance and reply tone (both opt-in and per-model overridable), and long-term memory gated by human confirmation - the AI can only append proposed candidates, a human confirm line flips them into effect, and the log is physically append-only and replayable. Confirmed entries are injected as a quoted data-not-instructions block; entries tagged with a kind can instead be routed to arbitrary files on confirmation and never reach the prompt. Three prompt sections re-evaluate from disk on every assembly, so edits apply on the next step with no restart. Persona presets are shareable JSON snapshots (imports and exports SillyTavern character cards, v2/v3 read, v2 written) and deliberately never carry memory. Zero runtime dependencies, no network, no shell; any error degrades to an empty section rather than a broken session. Ships blank - installing changes nothing until you write something.

Usage and cost view for DeepSeek Harness: aggregates real token usage from existing session logs, locks each cost at the price of its event time, and shows month and today spend against a budget from a sidebar card or an input-bar pill.

Feishu/Lark gateway for DeepSeek Harness with topic-persistent conversations, live execution cards, image and file inputs, in-chat approvals and questions, source allowlists, and durable one-time or recurring agent tasks created in natural language.

Spec-Driven Development pipeline for Odoo: 15 tools take a request from an immutable spec through architecture to module code or live configuration, then require a persisted verdict backed by evidence from a real Odoo instance, with fail-closed gates, human-approved batches in a declared environment, Odoo's own CSV/Excel importer, and a journal that reverses the data written through the plugin.

FriendliAI serverless LLM adapter for DeepSeek Harness — OpenAI-compatible chat completions with dynamic model discovery and reasoning support

Free Kilo Gateway and OpenCode Zen models for DeepSeek Harness (DSH) via native dsh-llm adapters.

Zero-config lint feedback loop for JS/TS, Python, Go and Rust: lint_diagnostics / lint_workspace_errors / lint_fix tools driven by the repo's own eslint / biome / ruff / golangci-lint / cargo clippy (auto-detected from config files, repo-local node_modules/.bin resolved first, nothing bundled), plus an auto-injected post-edit findings delta on the harness fs/observed event and a bounded completion gate — edit → lint → auto-repair with one lint_fix call for dsh agents.

Cross-session memory with governance, self-evolution and an approval-gated write path for DeepSeek Harness. Heat-ranked recall, environment conventions, forgetting lifecycle, supersession chains, poisoning defense, auto-distillation, and an open measureme

Freshness gate for the agent's test, typecheck, lint and build claims: each verification run becomes a receipt on tools/post-execute, file edits are tracked, and on agent/turn-stopping a claim whose latest receipt is stale, failed or has no readable result steers one more step naming the receipt and the files edited since.

Six-stage WeChat Official Account content-publishing workflow with quality-gate tools and a bundled skill.

Agent discipline guard for DeepSeek Harness: four gates enforced at the harness boundary — loop circuit-breaker, token-cost fuse, route-switch warning, plan approval — where prompt-level reminders can't reach, so an agent can no longer burn a weekly quota in minutes; one-click 🛡 chip, hot-reloadable thresholds, and presets from invisible-but-armored to warn-only.

DeepSeek Harness peak/off-peak pricing gate: during peak hours (weekdays 09:00-12:00 / 14:00-18:00 Beijing time) every send asks for confirmation, with an option to mute the reminder for the current peak segment, or to queue the message for off-peak (half

WeChat official account AI writing pipeline: six-step article runs (topic, outline, draft, gates, render, images) with a launch brief contract, a hot-topic board with per-item AI digests, RRULE scheduling to the draft box, a workbench UI plus chat tools and a /wewrite command, and fail-closed host approval before any draft-box push.

Agent preset that turns multi-stage objectives into resumable flows: refereed stages, SHA-256 lock identity, an HMAC-signed terminal review gate and event-sourced run logs; ships 15 agent tools.

Compaction fix for NInfer-hosted qwen3.8-27b gateways: disables thinking on dsh context-compaction and session-title calls so xhigh reasoning stops burning the whole output budget; the same idea applies to other launch methods, this package targets NInfer only.

Requires one-shot user approval before configured tools (default bash, pwsh) execute — gated tools pause and ask, everything else delegates, and a missing approval channel fails closed.

Config-driven development-task orchestration for DSH: graph workflows (task/gateway/human nodes) with per-stage persistence, plus an in-editor AI assistant that configures workflows through conversation with live validation. Hand documents to nodes as generic attachments (no declaration needed): files land in the task workspace and ${inputs.附件} gives the newline-joined absolute paths, so a node parses them with its own prompt.