Copilot Config Promptops

Build and maintain a persistent, compounding LLM wiki — a structured collection of interlinked markdown pages that sits between you and raw sources. Implements the Karpathy LLM Wiki pattern: the LLM incrementally reads sources, integrates knowledge into wiki pages (not re-derives it on every query), and files answers back so knowledge compounds over time. Three core operations: INGEST a new source into the wiki, QUERY the wiki and file the answer back, LINT the wiki for health. Integrates with Zettelkasten MCP for atomic cross-domain notes, plus web search, Semantic Scholar, and the workspace's existing log.md and inbox. Distinct from wiki-builder (which scaffolds wikis) — this skill is the ongoing methodology: how to operate an LLM wiki once it exists. Use when: "build an LLM wiki on", "ingest this source into my wiki", "query my wiki about", "lint my wiki", "set up a wiki for [topic]", "add this to my wiki", "file this answer", "wiki is getting stale", "expand my wiki", "start a knowledge base that compound

Store and retrieve bookmarks in Raindrop.io using the Raindrop MCP server. Use when saving research sources, PDFs, articles, or any URL encountered during work. Also use when searching or browsing existing bookmarks by tag, text, or collection.

Build and maintain topic-scoped "LLM knowledge bases" — structured agent-maintained wikis for deep-diving a single topic (paper, framework, person, project, domain). Scaffolds a folder with wiki.config.md, raw/, prompts, and sources.md in one command. Compiles raw source material into organized wiki pages, files Q&A answers back into the wiki, and runs maintenance passes to find thin pages and missing links. Works standalone (flat markdown) or integrated with zettelkasten-mcp (compiled pages become ZK notes tagged with the wiki topic). Inspired by DAIR.AI's Wiki Builder plugin, adapted for VS Code + GitHub Copilot. Triggers: "start a wiki on", "build a knowledge base about", "create topic wiki", "wiki for this paper", "deep dive into X", "research wiki", "compile this into a wiki", "knowledge base for project", "file this answer", "lint my wiki", "wiki-builder".

You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

Multi-tiered deep research methodology — from raw data gathering through evaluation to synthesis. Use when conducting comprehensive research on any topic, building evidence-based arguments, or investigating complex questions.

Audits, scores, and rewrites AI configuration files to maximize token efficiency with zero behavior loss. Use when asked to optimize, compress, or reduce tokens in copilot-instructions.md, AGENTS.md, .github/instructions/*.md, .github/copilot-instructions.md, .github/agents/*.agent.md, skills/*/SKILL.md, or any system prompt / AI instruction file. Runs a 4-phase workflow: inventory → waste detection → rewrite → measure. Triggers on: "reduce tokens", "optimize instructions", "compress prompt", "too verbose", "token budget", "shrink system prompt", "refactor copilot config", "clean up instructions".

Build and manage a networked knowledge base (second brain) using the Zettelkasten method with the zettelkasten-mcp MCP server. Use when creating atomic notes, linking ideas, searching the knowledge graph, building Maps of Content (MOCs), capturing research, processing meeting notes, or running knowledge maintenance. Includes installation from the GitHub repo via uvx, mcp.json configuration, full MCP tool reference, note type taxonomy, semantic link types, and Luhmann's methodology. Triggers: "zettelkasten", "second brain", "atomic note", "knowledge graph", "create note", "link ideas", "zk_create_note", "install zettelkasten-mcp", "MCP config", "note taking", "Luhmann", "slip box", "build knowledge base", "literature note", "permanent note".

Academic literature review using Raul Pacheco-Vega's methods (AIC reading, Conceptual Synthesis Excel, research memos). Use for systematic reading and synthesis of academic or professional literature on a topic.

Use when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions

Always-on context hygiene. Reduces token waste by using rtk (Rust Token Killer) as a CLI proxy for all terminal commands and offloading memory to external MCPs. Apply to every session — no exceptions. Triggers: any terminal command, any task that reads files or runs shell commands, any research task that would keep large outputs in-context.

Progressive VS Code + GitHub Copilot setup coach. Interviews the user and/or scans their workspace to determine their current maturity level, then guides them through one small, high-value improvement at a time. Covers the full stack: Copilot license, extensions, global and scoped instructions, prompts, custom agents, skills, MCP servers, and governance workflows. Maintains session continuity via Zettelkasten notes between conversations. Use when starting a new setup, stuck on what to configure next, onboarding a colleague, or doing a periodic health check. Triggers: "help me set up Copilot", "what should I configure next", "copilot setup", "onboard me to Copilot", "improve my AI setup", "set up GitHub Copilot", "what MCP should I install", "show me what to do next", "copilot health check", "audit my Copilot setup", "next improvement".

Apply 180+ thinking frameworks to any problem. Use when analyzing, deciding, strategizing, brainstorming, evaluating evidence, designing solutions, or coaching others through structured thinking. Covers mental models, decision-making, strategy, creativity, systems thinking, cognitive biases, and more — organized into 15 categories with guidance on when to use each. Do not use when a direct factual answer suffices.

Manage secrets with 1Password CLI - retrieve credentials, create items, enforce naming conventions, integrate with Docker/CI/CD. Use when user asks to "get secrets from 1Password", "create 1Password item", "standardize vault names", "integrate 1Password with Docker/scripts", or "query op vault".

Use when deploying or operating self-hosted Authentik as infrastructure, especially through Ansible playbooks, Docker Compose, Traefik, PostgreSQL, Redis, backups, upgrades, outposts, and production service validation. Prefer this skill when the user wants to build or run Authentik automation on servers, template compose or env files, manage host-level integrations, or handle operational changes beyond documentation lookup.

Shared knowledge base for the dk.speckit.* agent family. Defines evidence-first execution, risk classification, pushback protocol, verification strategy, and artifact contracts. Agents reference this skill to stay DRY — philosophy and formats live here, agent files focus on workflow.

Use when the task depends on the self-hosted Authentik community documentation, especially docs.goauthentik.io pages about blueprints, managed/blueprints, export_blueprint, Docker Compose, .env settings, manual outposts, and OSS-only configuration patterns. Prefer this skill when the user wants Authentik guidance grounded in upstream docs, blueprint schema details, or explicit exclusion of Enterprise features.

Planning-only skill for data analysis projects. Use when starting any data analysis project, notebook workflow, R Markdown/Quarto analysis, Python analysis, statistical study, or AI-assisted analysis repo before writing analysis code. Produces a token-efficient PLAN.md, atomic analysis-unit structure, reproducible environment plan, idempotent setup/validation/bootstrap scripts, caching/provenance strategy, and handoff to specialist analysis skills. Triggers on planning data analysis, setting up analysis projects, structuring notebooks/Rmd files, avoiding giant notebooks, keeping calculation and interpretation together, saveRDS/cache/targets/provenance, or Copilot-friendly analysis workflows.

Comprehensive information architecture (IA) skill for designing, evaluating, and implementing information structures across digital products, knowledge management systems, and scientific domains. Covers IA design process, card sorting, tree testing, PARA, Zettelkasten, Hub-and-Spoke architecture, taxonomy/ontology design, and knowledge graph patterns. Use when asked to 'design an IA', 'structure information', 'organize content', 'card sort', 'tree test', 'create taxonomy', 'build a knowledge system', or any task involving information structure, navigation design, or knowledge organization.

Generate, validate, and deploy LimeSurvey surveys using a markdown-first pipeline. Use when asked to create, edit, or deploy a LimeSurvey survey as a markdown file, when running `limesurvey validate`, `limesurvey convert`, or `limesurvey api` CLI commands. Covers markdown schema, question types, relevance/branching logic, parser quirks, and the full validate→convert→deploy workflow.

Orchestration logic for Zettelkasten knowledge management — routing, lifecycle management, maintenance protocols, and health checks. Use when managing the Zettelkasten system at scale, processing backlogs, making maintenance decisions, or coordinating between research and note-taking agents.

Diagnose a machine and set up a developer environment with VS Code, GitHub Copilot, Node.js, and Python. Supports use-case-specific tracks: data science (Python/R), AI-assisted development, web development, and general programming. Installs packages per use case with reporting from Markdown to Quarto. Use when: setting up a new machine, onboarding a developer, bootstrapping a dev environment, data science environment, R setup, Python data science, Quarto, Jupyter, AI dev environment, checking what is installed, detecting package managers, checking user permissions, OS detection, Windows/Mac/Linux setup, environment audit.

Build, review, refactor, and operationalize a GitHub Copilot customization control plane in a repository. Use when users ask how `.github/copilot-instructions.md`, `.github/instructions/*.instructions.md`, `.github/prompts/*.prompt.md`, `.github/agents/*.agent.md`, and `.github/workflows/*.yml` should interact, or when they ask to create or update these files directly (for example: "create a new agent", "save this as prompt", "add scoped instructions", "add a governance workflow", "create new prompt", "create new instructions file", "edit prompt", "edit instructions file").