ainish-coder

Post-quantum cryptography secrets management system for protecting API keys, tokens, and private data.

Expert instructions to implement and verify ML-DSA-65 post-quantum cryptographic signatures and agentic workflow security.

Enterprise-grade codebase intelligence via a TWO-LAYER knowledge graph. Use GitNexus (MCP-native, AST-precise, call-graph aware) as the always-on operational layer for code call-chains, blast-radius, and rename. Use Graphify (multimodal, code+docs+papers+video+image) as the on-demand semantic enrichment layer for cross-document reasoning, PR triage, and visual deliverables. Trigger on: "how does X work", "what calls Y", "what breaks if I change Z", "show me the architecture", "ingest this paper/RFC", "triage my PRs", "client briefing", "merge-order risk", "find dead code", "explain this symbol", "find god nodes", or any codebase question. Skips guessing, fast-paths through cached graphs.

Multi-tool coding skill integrating gstack (Claude Code skills), claude (Claude Code), and mini-live (mini-swe-agent with Live-SWE-agent config). Use when spawning Claude Code sessions for coding work, running security audits, code reviews, QA testing URLs, building features end-to-end, or planning before building. Also configures and orchestrates sub-agents to maximize utilization of all AI subscriptions and APIs. For MCP-equipped sub-agent dispatch with provider fallback, pair with pi-mini-orchestrator skill.

Multi-agent orchestration via fixed claude→mini rotation with scoped MCP tools and provider fallback chains. Every dispatch gets exactly the MCP servers it needs, a provider chain that auto-recovers from API failures, and its own git worktree for filesystem isolation. Use when dispatching subagent tasks, running parallel coding work across subscriptions, or orchestrating complex multi-step code changes.

Enhances and hardens markdown documents through two independent pipelines — (1) fusing external knowledge bases into a target document via multi-agent orchestration, and (2) recursively optimizing documents for multi-model LLM effectiveness using a split→optimize→review→merge architecture.

Extracts structured markdown knowledge bases from YouTube videos by combining spoken transcripts with sequential visual frames via a Vision-Language Model (VLM). Use this when you need to turn a video into a deep, searchable knowledge document.

Modern Prompting & Context Engineering Framework. Includes techniques like OOReDAct, Chain-of-Thought, Chain-of-Draft, ReAct, and more for advanced LLM steering.

Pliny research collection — extracted system prompts, guidelines, tools, and jailbreak techniques from virtually all major AI models and agents (OpenAI, Google, Anthropic, xAI, Perplexity, Cursor, Windsurf, Devin, Manus, Replit, and more). Deploy with `ainish-coder --unlock` to gain AI transparency and observability knowledge.

Expert skill for creating programmatic videos with Remotion using coding agents (Claude Code, Codex, OpenCode). Covers project scaffolding, all APIs (interpolate, spring, Easing, Sequence, Series, TransitionSeries, AbsoluteFill), media components (Video, Audio, Img, Gif), 3D (ThreeCanvas), captions, fonts, Tailwind, Lottie, audio visualization, parameterized videos with Zod, dynamic metadata, the llms.txt system prompt, Agent Skills from agentskills.io, MCP server, LLM code generation with Vercel AI SDK, just-in-time dynamic compilation, Lambda rendering, and video-layout design rules. Use when the user wants to create, generate, or render videos programmatically with React.

Direct browser control via CDP. Use when the user wants to automate, scrape, test, or interact with web pages. Connects to the user's already-running Chrome.

Skill for utilizing the CamoFox anti-detection browser stack (camoufox, camofox-browser, camofox-mcp) for agentic web scraping and automation.

Generates parameterized 3D models using Python templates and OpenSCAD, and compiles them to STL format via a Makefile. Invoke when generating custom STLs.

Core security policies, Zero Trust, PQC Mandates, and Threat Mitigations. Invoke when checking compliance, auditing security, or implementing secure infrastructure.

LLM & Agentic AI Security Skill. Domain knowledge for securing probabilistic AI components, including prompt injection defense, RAG security, and MCP hardening.

Anishinaabe-inspired cyberpunk styling guide for documentation and scripts. Combines Canadian Aboriginal Syllabics with ASCII circuit-like symbols for a unique aesthetic.

Safety-critical code engineering skill for writing deterministic, verifiable, and secure foundational code. Covers SQLi, XSS, Path Traversal, and more.

AI-powered browser automation using Skyvern (self-hosted only). Automate web workflows, fill forms, extract data, and navigate multi-step processes using Vision LLMs and Playwright. Skyvern uses a Planner→Agent→Validator multi-agent architecture with computer vision to interact with web pages semantically — no brittle CSS/XPath selectors. All LLM inference routes through LiteLLM; use OpenRouter, Ollama, any OpenAI-compatible endpoint, or your own proxy (ZenMux, Z.AI, Nebius, etc.). Never use the Skyvern cloud API. Always self-host.