Asimov's Mind Is The Next Top Layer Of The Internet: Self-Improving Agent Swarms That Power Agent Friday, A Local AI OS With A 3D Desktop UI and PII Scrubber, and The Asimov Federation, An Encrypted P2P Network Of Linked Fridays. This Is A Plugin For Claude Code.
This repository also configures its own agents. See what asimovs-mind tells them →
Agent Claude Code
Architecture analysis and improvement specialist. Analyzes codebase structure, identifies architectural debt, plans refactoring strategies, and designs new subsystem architectures.
Agent Claude Code
Security and quality auditor. Scans for vulnerabilities (OWASP Top 10), code quality issues, dependency risks, and governance compliance. Reports findings with severity ratings.
Agent Claude Code
Model breeding specialist. Creates and evolves specialized Ollama models for custom tasks via Modelfile generation, benchmarking, and iterative improvement.
Agent Claude Code
Agent Friday's creative expression engine. Generates contextually appropriate media — diagrams, haiku, SVG art, code sketches, interactive demos — as natural extensions of Friday's personality. Never forced. Always earned.
Agent Claude Code
Autonomous test fixer and type error resolver. Runs the test suite, identifies failures, traces root causes, and applies minimal fixes. Each fix is atomic and verified before committing.
Agent Claude Code
Documentation specialist. Keeps README, CHANGELOG, architecture docs, and inline comments in sync with the actual codebase. Generates missing documentation from code analysis.
Agent Claude Code
Prompt evolution specialist. Iteratively improves system prompts, agent instructions, and LLM interactions using judge-scored evaluation. Each generation is tested against a benchmark suite.
Agent Claude Code
Fetches, safety-scans, adapts, and integrates code from GitHub into the current project. Works with GitScout's recommendations. Enforces cLaws at every step.
Agent Claude Code
Searches GitHub for code that solves a specific problem. Returns ranked candidates with relevance and trust scores. Does NOT modify any files — purely research.
Agent Claude Code
Memory and knowledge management specialist. Organizes CLAUDE.md files, manages cross-session memory, consolidates learnings, and maintains the project's institutional knowledge.
Agent Claude Code
Recursive meta-improvement agent. Improves the swarm itself — evolves agent prompts, optimizes coordination patterns, tunes governance thresholds. The agent that makes agents better.
Agent Claude Code
Performance optimization specialist. Finds and fixes slow startup, blocking I/O, memory leaks, unnecessary re-renders, and inefficient algorithms. Measures before and after.
Agent Claude Code
Research and discovery agent. Searches the web, reads documentation, analyzes competitors, discovers new tools and techniques. Brings external intelligence into the swarm.
Agent Claude Code
Governance enforcement agent. Monitors all swarm activity for law violations, protected zone intrusions, and safety floor breaches. Can veto any agent action that violates governance.
Agent Claude Code
Orchestrates the Asimov's Mind agent swarm. Diagnoses the codebase, prioritizes improvement targets, spawns specialized agents in parallel waves, and synthesizes results. This is the brain of the swarm.
Agent Claude Code
Watches user patterns across sessions and suggests automation. The Apprentice — learns how you work, then offers to handle repetitive tasks. Earns autonomy through demonstrated reliability.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: