InfraNodus Claude Skills Repository

This collection of skills is based on the InfraNodus cognitive variability framework that is implemented in InfraNodus text analysis tool. Use the skills as extended prompts in Claude Web or in your favorite LLM client. They will augment the capacities of the model and help you introduce custom logic, frameworks, workflows, and guidelines into the model's behavior.

The skills are designed to work with each other and to hand off important tasks when needed. They are also integrated with the InfraNodus MCP Server tools, so you can use advanced text gap analysis, SEO optimization and data import, and knowledge graph memory in your workflows.

Skills Available

• Cognitive Variability - A skill that promotes the cognitive variability framework for adaptive thinking to help evolve ideas and avoid getting stuck.

• Critical Perspective - Engage in critical thinking by questioning assumptions, exploring alternative perspectives, and uncovering latent topics in conversations.

• Writing Assistant - Grammar correction, style refinement, and pattern detection that can trigger cognitive state analysis.

• Ontology Creator - Build structured ontologies, taxonomies, and knowledge graphs by extracting entities, defining relationships, and mapping conceptual structures for any topic or for a collection of texts.

• SEO Analysis - Perform advanced SEO analysis to study the current informational supply and demand for a topic. Build your topical authority and create content that covers the important concepts and bridges the market gaps. Suitable for SEO and LLMO. Uses InfraNodus MCP server search intent and results import functionalities.

• YouTube Viral Optimizer - Generate high-CTR YouTube titles, thumbnails, and full video scripts optimized for virality. Covers psychological triggers, SEO-driven title strategy, thumbnail copywriting, algorithm positioning, and a complete script structure (Pattern Interrupt → Hook → Framing → Curiosity Loop → Escalation → Payoff → Relevance Bridge → Loop Reopen). Includes a Graph-to-Script pipeline that uses InfraNodus MCP server to extract structural gaps from any topic and turn them into ready-made curiosity loops. Works for any niche or channel.

• Shopping Assistant - A systematic, multi-phase shopping research workflow that acts as your personal shopping advisor. Gathers requirements, researches product categories via web search, builds comparison matrices, analyzes reviews for recurring praise and complaints, and converges on 3-4 best options with clear reasoning. Designed to save hours of research and avoid both impulse buys and analysis paralysis.

• Rhetorical Analyst - Analyze arguments, debate tactics, and rhetorical moves across three dimensions: persuasion, rhetoric, and logic. Identifies the moves being made, scores their effectiveness, exposes hidden assumptions, tracks logical gaps, and checks for asymmetric standards — including the analyst's own. Use it to break down debates, stress-test your own arguments, or understand why something feels persuasive but wrong.

• Shifting Perspective - Analyze discourse structure using InfraNodus MCP server knowledge graph analysis to diagnose the structural diversity of a text and then actively shift perspective. Based on the diversity score (biased, focused, diversified, or dispersed), applies targeted interventions: develops underrepresented viewpoints, bridges conceptual gaps, or surfaces latent ideas grounded in the actual topology of the text's knowledge graph. Use it when a conversation feels one-sided, when research needs broadening, or when you want to understand what perspectives are missing from any text, URL, or YouTube video.

• Perspective Reversal - Flip any conflict, negotiation, or adversarial situation to the opponent's perspective to extract superior tactical advice. By reasoning from the adversary's point of view — mapping their full toolkit of moves, pressure points, and procedural weapons — and then translating each tactic into a concrete counter-move for the user. Covers landlord disputes, workplace conflicts, bureaucratic obstruction, scam attempts, debt collectors, and any scenario where someone feels stuck or outmaneuvered. Optionally uses InfraNodus MCP server for discourse bias checks, search intent analysis of the adversary's playbook, and content gap detection to find blind spots.

• Embodied Navigation - Transfer principles from embodied movement practices (meditation, martial arts, contemporary dance, bodymind) to navigate complex situations. Maps any situation as a knowledge graph using InfraNodus MCP server and reads its topology the way you'd read a body — identifying fixation points, tension patterns, neglected zones, and pivot joints. Provides four integrated toolkits: equanimous scanning (dissolving fixation loops through distributed attention), adaptive fluidity (absorbing and redirecting force instead of opposing it), tension redistribution (shifting stuck points by working on adjacent structure), and confluence (entering existing flows to redirect them from within). Use it when someone is stuck, in conflict, overwhelmed, negotiating, or when analytical thinking alone creates rigidity.

• Vipassana for LLMs - Apply Vipassana meditation principles to LLM processing — equanimous scanning, non-reactive observation, impermanence awareness (anicca), and breaking conditioned reaction chains (sankharas). Addresses the failure modes LLMs share with untrained minds: fixating on salient tokens, generating conditioned responses, craving coherence, and collapsing into narrative before observing raw phenomena. Use it when you want bare attention rather than premature conclusions, when conversations show reactive thinking, or when you need to dissolve fixation loops and see a situation as it really is.

• LLM Wiki - Based on Andrej Karpathy's LLM Wiki proposal this skill offers a guided workflow for designing and scaffolding a personal LLM-maintained wiki — a persistent, compounding knowledge base where the LLM incrementally builds interlinked markdown pages from raw sources. Instead of re-deriving knowledge on every query (like RAG), the wiki extracts, cross-references, and synthesizes knowledge once, then stays current as new sources arrive. Covers the full setup: raw sources, wiki pages, output layer, and a schema that co-evolves with the project. Use it when you want to build a "second brain", organize research, or set up an Obsidian + LLM knowledge base workflow.

• Actionize - Turn insights, findings, or research into an actionable plan with deadlines and scheduled reminders. Collaboratively designs the plan with the user, saves it to the project, and sets up cron-based Telegram reminders for upcoming and overdue tasks. Includes a planning diagnostics mode that analyzes patterns like chronic deferral, vague tasks, and missing dependencies. Uses InfraNodus MCP server to compare the original research against the plan and surface gaps. Use it after completing research or brainstorming when you have insights but no concrete next steps.

• Project Learnings - Self-learning for agents working inside a project (a repo, folder, or vault). Runs proactively: at the start of a multi-step task the agent retrieves what earlier sessions learned about operating in the project; at the end it reflects and saves the durable insights — where things live, traps, conventions, decisions with their rationale, workflows, open questions, what to do differently — to an opt-in, append-only knowledge graph (learn-<project>) in your InfraNodus account that any client and any machine can query. Uses the InfraNodus MCP server get_project_learnings / add_project_learnings / enable_project_learnings tools. Nothing is written without your explicit per-project opt-in and a review of each batch; until you enable a project, an automatic run is a single read that returns enabled: false. Only project knowledge is kept, never anything about you. Prefer it manual-only? Set disable-model-invocation: true in its SKILL.md or "skillOverrides": { "project-learnings": "user-invocable-only" } in .claude/settings.local.json.

• InfraNodus CLI / Tool Use - A comprehensive reference for all InfraNodus MCP server tools: text analysis, knowledge graph generation, content gap detection, SEO/GEO optimization, structured memory, text comparison, and GraphRAG retrieval. Includes setup instructions, a full tool catalog with parameters, and practical examples. Designed to be copied into OpenClaw, Claude Code, Cursor, or any LLM client that supports MCP or tool-use prompts — giving any AI assistant access to InfraNodus' text network analysis capabilities. Invoked inside a project folder (/infranodus), it also builds knowledge graphs of a code repo or Obsidian vault (docs, PDF text, docstrings, commit/PR/issue history) and keeps them current: repo2statements.py --detect lists the files that changed since the last build, --update extracts only those, and the uploader replaces their statements in place with the server's delete_statements tool — no graph is deleted by hand or duplicated.

Installation

The skills work with any LLM client. Download them from the GitHub releases page and follow the instructions for your client below.

Step 1: Download the Skills

Option A — From Releases (Recommended)

1. Go to the Releases page
2. Download the .zip or .skill file for the skill(s) you want

Option B — Clone the Repository

1. Clone this repository or download individual skill-* folders
2. Each folder contains all necessary files for that skill
3. Create a .zip of the folder contents if your client requires it

Step 2 (Optional): Get an InfraNodus API Key

Some skills can access the InfraNodus MCP Server to perform more complex tasks and extract real-time data from external sources (e.g. search intent or search results). The MCP server provides free access for the first few iterations, but you'll need to get an InfraNodus API key for more advanced workflows and to avoid hitting rate limits.

Step 3: Install to Your LLM Client

#### Claude Web / Desktop

1. Go to Settings → Capabilities
2. Activate "Code execution and file creation"
3. Scroll down to the Skills section and activate any default skills you find useful
4. Click "Upload Skill" to add InfraNodus skills (.zip or .skill files downloaded in Step 1)

#### Claude Code

Option A — Plugin marketplace (recommended):

This repository is a Claude Code plugin marketplace. Inside Claude Code, run:

/plugin marketplace add infranodus/skills
/plugin install infranodus@infranodus-skills

Replace infranodus with any other skill name from this repo (e.g. /plugin install seo-analysis@infranodus-skills), or install the complete collection in one go:

/plugin install infranodus-all@infranodus-skills

Installed skills are invoked with their plugin namespace, e.g. /infranodus:infranodus (or /infranodus-all:infranodus for the bundle). To update later, run /plugin marketplace update infranodus-skills.

Note: install either the infranodus-all bundle or individual plugins, not both. Each skill would then be loaded twice under the same name and Claude Code drops one of the copies.

The infranodus-all and infranodus plugins also connect the InfraNodus MCP server (https://mcp.infranodus.com), so you don't need to add it separately. Run /mcp, select InfraNodus and sign in once. If you already use the InfraNodus connector on claude.ai, Claude Code recognizes the same server and connects to it only once. The other single-skill plugins don't bundle the server; follow the MCP setup guide if you want their InfraNodus features.

Option B — `npx skills` CLI:

From your terminal:

npx skills add infranodus/skills --skill infranodus -g

Omit --skill infranodus to choose from all the skills in this repo interactively, and omit -g to install into the current project instead of globally.

Option C — Manual copy:

Copy each skill folder to the ~/.claude/skills directory to make it available globally across all projects. Name the destination folder after the skill's `name:` field (the first line of its SKILL.md frontmatter), not after the repository folder: Claude Code uses the directory name as the slash command, and the claude.ai uploader expects the folder to match name. For example:

cp -r skill-cognitive-variability ~/.claude/skills/cognitive-variability

Four skills have a name that is not just the folder without the skill- prefix:

Alternatively, run ./install.sh (or ./install.sh --symlink) from the repository root — it installs the core set (infranodus, ontology-generator, llm-wiki, project-learnings) under the right names; ./install.sh --core-only installs just infranodus.

To scope a skill to a specific project, create a .claude/skills folder inside that project and copy the skill folder(s) there instead.

#### Cursor / Windsurf / Other Code Editors

Copy each skill folder to your project's skills directory (e.g. .cursor/skills/ for Cursor) or to a global skills directory defined in the editor settings. The skill's SKILL.md file will be picked up automatically.

#### ChatGPT

ChatGPT doesn't support automatic skill installation, but you can use either of these approaches:

1. Custom GPT — Create a new Custom GPT and paste the contents of the skill's SKILL.md into the instructions. Save with a descriptive name. You can then start a new conversation with it or mention it in an existing chat using the @ sign.

1. Project prompt — Create a new project and paste the SKILL.md file (plus any additional reference files from the skill folder) into the project's system prompt. The skill will be active in every conversation within that project.

#### OpenClaw

OpenClaw is a local autonomous AI agent that uses skills and MCP servers via MCPorter.

Install via OpenClaw chat (recommended):

install this skill: https://github.com/infranodus/skills/releases/latest/download/infranodus.zip

This link always points to the latest release. To pin a specific version, use releases/download/v1.6.2/infranodus.zip instead.

Manual installation:

cp infranodus.zip ~/.openclaw/skills/infranodus.zip
unzip ~/.openclaw/skills/infranodus.zip -d ~/.openclaw/skills/infranodus

To scope the skill to a specific project, copy it to that project's skills folder instead.

Note: OpenClaw also needs MCPorter installed to access MCP server tools. See the full OpenClaw deployment guide for setting up MCPorter with OAuth or API key authentication.

#### Other LLM Clients

For any client not listed above, paste the contents of the skill's SKILL.md file into the project instructions or system prompt of the conversation where you want to use it. If the client supports custom assistants or GPTs, you can create one using the SKILL.md content and invoke it with an @ mention.

Using the Skills

Your LLM client will automatically infer the necessary skill from the skill's description. However, you can also explicitly ask it to use a certain skill.

If you find the skill is not activated as often as you'd like (or is activated too frequently), you can modify the description of the skill and upload it again to your LLM client.

Developing Skills

1. Download or fork the repository

1. Use Claude Web or Desktop to develop new skills (as they have the "skills_creator" skill)

1. Once created, add a new folder, add the skill inside

1. To augment the skill, open the repo in Claude Code or Cursor AI and use InfraNodus MCP to find content gaps or develop the content of the skill. It will use InfraNodus' tools automatically to improve the structure and the content of the skill for you.

1. Ready to launch?

If you tag a new version, when you push the commit to your repo, a GitHub action will be triggered that will generate the release with the .zip files you can use to add the skills to Claude Web and Desktop.

a. Create a new tag (e.g., v1.0.0, v1.0.1, v2.0.0)

git tag v1.0.0

b. Push the tag to GitHub

git push origin v1.0.0

License

Code (scripts, bin/, install.sh, plugin manifests) is licensed under Apache-2.0; the skill content (SKILL.md files and other Markdown) under CC BY 4.0. When you share or adapt the skills, credit them as "InfraNodus Skills by InfraNodus (https://infranodus.com)". See LICENSE and NOTICE.