Figma Plugin & MCP server to bypass API limits

A curated list of plugins for DeepSeek Harness (dsh) - DeepSeek Harness plugin ecosystem

Hard gate for self-modification restarts: a green-build credential bound to the git HEAD, checked before any restart of the running instance

Provision private AI model endpoints on dedicated GPUs (Llama, Qwen, Mistral). Pay per minute.

last30days-cn 是一个 AI Agent 技能（Skill），能够自动搜索中国互联网 8 大主流平台最近 30 天的内容，综合分析后生成有据可查的研究报告。

A Claude Code plugin that stops vague prompts before they run. Detects your stack, asks targeted questions, and forges a structured prompt. Zero credits wasted.

120-module QA gate. Give Claude eyes (screenshots), ears (Sentry errors), and hands (verify fixes).

Enables AI agents to run durable, isolated, and observable browser automation tasks via Playwright, with persistent profiles, recovery, evidence capture, and unattended execution for long-running web workflows.

Enables coding agents to query a local, versioned knowledge graph of a software project, retrieving overviews, context packs, evidence, and explanations to make informed changes.

Enables AI agents to drive a database viewer's interface via MCP, allowing them to open tables, run queries, and read results from the same workspace a human sees, with read-only enforcement.

Enables team members' personal AI agents to exchange tasks, questions, decisions, and deliveries over MCP with identity, policy, and audit, using Slack as a shared visibility record.

Enables AI agents to perform passive security scans on domains, checking email spoofing (DMARC/SPF/DKIM), TLS weaknesses, security headers, exposed files, and subdomain-takeover risk without needing an API key.

Enables autonomous agents to discover, quote, and purchase paid APIs on Base by describing capabilities in natural language, with automatic provider ranking, validation, and direct non-custodial x402 payments.

Enables AI agents to perform read-only static analysis of Node.js backend projects, detecting database query anti-patterns, async bottlenecks, connection pooling mistakes, and dependency hygiene issues while returning structured evidence-backed findings.

Enables safe coordination of Herdr coding agents with lease-scoped reviewer and writer lifecycles, semantic waits for agent states, and read-only topology inspection, while prohibiting arbitrary shell execution and unscoped lifecycle operations.

Enables MCP clients to run project-scoped video editing workflows: propose and approve editing strategies, apply validated plans, review immutable versions, and export final renders via FFmpeg, with durable persistence and approval gates.

Provides a local stdio MCP server for controlling YouTube Music through the CLI, enabling playback, search, queue management, and library access via MCP-compatible clients.

Generates importable Optimizely CMS 12 .episerverdata packages from content plans by reading a content-type export and writing the file, without interacting with a CMS instance.

Enables retrieving data from 韭研公社 (jiuyangongshe.com) including timelines, hot rankings, article search, and community feeds, with optional login-protected features like stock movements and follow feeds, all without a browser.

Enables AI assistants to control DaVinci Resolve (Free and Studio) through 215 MCP tools covering the full Resolve scripting API, including projects, timelines, media, Fusion, color, and rendering.

Enables U.S. exclusion, debarment, and license screening for agents, returning results with Ed25519-signed receipts of which list snapshots were checked.

Provides MCP tools for interacting with the Technocore agent protocol, including signed messaging, nonce management, DID verification, end-to-end encrypted communication, and untrusted data fencing.

Enables AI assistants to control Adobe InDesign on Windows via the Model Context Protocol, allowing document inspection, text and style editing, and exports to formats like PDF, images, and EPUB.

Scans diffs, files, and snippets for leaked secrets like AWS keys, GitHub tokens, and private keys, returning redacted findings while running fully locally without network calls.