Browse agent skills

A Claude-Code-shaped permission rule table as pre_tool_use / on_resume stages.

Local-first LLM gateway aggregating free/low-quota model APIs for coding agents

Client SDK for talking to a Laintern chat agent: token lifecycle, streaming, citations and sources.

Slack tools for DeepSeek Harness: conversations, members, rich messages, scheduled messages, user groups, file metadata, reactions, and notifications

Drive the Proxya Anty anti-detect browser from Node. Fingerprints are applied inside a patched Chromium, and the DevTools Runtime agent is not left switched on — so a script-driven session is not read as automated.

Local-first file conversion for DeepSeek Harness - images, PDF, data, audio/video and office documents. No API keys, no uploads, no token cost.

OpenCode TUI sidebar plugin showing real-time LLM metrics (TPS, TTFT, tokens) with sub-agent × model breakdown

DeepSeek Harness plugin: mini-app host wiring, panel, and skills

Reference instance-side client for the Nodo rail: keypair + one-time registration, signed check-in scheduler, local entitlement enforcement with offline grace.

36 data-engineering skills for AI coding agents — dbt, Airflow, Dagster, Prefect, Spark, Snowflake, BigQuery, Databricks, Kafka, Iceberg. Portable SKILL.md Agent Skills (the format Anthropic launched as Claude Skills), with idempotency, backfill, and data-quality know-how. Runs in Claude Code, Cursor, Codex, Copilot, Gemini CLI and more.

Rust-native file parser converting enterprise documents to LLM Markdown

Finch Agent CLI for wallet custody, Session-owned execution, and MCP OAuth approval.

Skill bundles for DeepSeek Harness: a bundle box holds a root skill and sub-skills, loaded by metadata (loadSubskills) or by model decision from body text.

Network & proxy assistant for DeepSeek Harness (dsh): one-shot GitHub connectivity check, system proxy status, proxy port probing, full diag chain (DNS/TCP/HTTP) and hosts-file conflict scan — built from real-world China-network experience.

Framework-agnostic core of the Tracepoint feedback & diagnostics SDK.

Accessible On-Demand YouTube Import, Live Streams, Smart Recommendations and A11y Engine for PeerTube

Reviews any draft for AI slop by writing where the point is fine but it's dressed up as a stock formula, forced metaphor, or filler instead of clear analysis. Flags it, explains why, and rewrites it in plain language. Works as a Claude Skill or a portable Gemini/ChatGPT system prompt.

Windows-native capability pack for DeepSeek Harness (dsh): clipboard, notifications, hosts file, network diagnostics and more — safe PowerShell-backed tools for Windows users.

Pi Task Monitor — always-on-top real-time task panel

NEXUS AI — Local-first AI workstation CLI. Think, plan, execute — powered by local AI runtime.

Compatibility patch for Better Sidebar app-frame spacing in DeepSeek Harness Plus

DeepSeek Harness 语音输入插件：在输入框加入麦克风按钮，用浏览器 Web Speech API 把语音实时转成文字填入输入框。

Community automation-only Agent Client Protocol server for DeepSeek Harness with committed tool-activity frames (tool_call/tool_call_update) — a drop-in alternative transport to @deepseek-ai/dsh-acp

Voice assistant plugin for dsh web: hands-free wake phrase, dictation, voice edit commands, and Chinese TTS.

Academic paper search, lookup, citation analysis, and PDF download for Pi coding agent. Searches Semantic Scholar, OpenAlex, Crossref, arXiv, DBLP, and PubMed. Supports Unpaywall for open-access PDF discovery.

Fold the DSH skill catalog prompt surface: static KV-cache-stable catalog + BM25/bge-m3 hybrid skill_search + autoRoute hints. Rendering-only fold: source.entries stay untouched, the skill tool and /name injection stay host-owned. v0.3.0 adds a semantic r

把本机登录的 WorkBuddy 模型接入 DeepSeek Harness，并提供只读的积分概览与模型管理。

并行多角色代码评审小队：/review 一条命令派出安全/正确性/测试/风格四名只读评审员子代理（每人可单独指定模型与思考强度），汇总为按严重度分组的结构化报告。Parallel multi-role code review squad: /review dispatches security/correctness/tests/style reviewer subagents (each on its own model) and aggregates one severity-grouped repo

Your agent runs the project through a verified loop: task → artifact (ADR / spec / epic / story) → adversarial critic → backlog → planner → reviewer → done. Plain markdown in git — any model can pick the project up tomorrow.

CodeGraph integration for DeepSeek Harness