Stats
Actions
Available In
Tags
Plugin
learning-system
Auto-improve, memory management, and second-brain workflows for the agent company.
What's Inside
Slash Commands12
ar:loop
/ar-loop
Start an autonomous experiment loop. The experiment-runner agent runs iterations continuously, keeping improvements and discarding failures, until interrupted or the goal is met.
ar:run
/ar-run
Run a single iteration of an autoresearch experiment. Reads state, decides one change, commits, evaluates, and keeps or discards.
ar:setup
/ar-setup
Set up a new autoresearch experiment interactively. Creates the .autoresearch/ directory structure with config, program.md, and optionally a starter evaluator.
ar:status
/ar-status
Show experiment dashboard — progress, best metric, keep rate, and results table for one or all experiments.
compile-raw
/compile-raw
Process all unprocessed files in the raw/ folder and compile them into canonical knowledge base pages. Handles mixed source types — markdown, PDFs, YouTube transcripts, Twitter exports, web captures, and more.
Agents4
Learning System Agents
/README
Specialized agents for the learning-system plugin.
experiment-runner
/experiment-runner
Autonomous experimenter that optimizes a target file by a measurable metric. Reads experiment state from `.autoresearch/`, decides what ONE change to try based on accumulated evidence and strategy escalation, makes the change, commits, runs the evaluator, and keeps or discards based on outcome. Spawned by `/ar:loop` and `/ar:run`.
memory-analyst
/memory-analyst
Read-only analyst for `~/.claude/projects/<project>/memory/`. Identifies promotion candidates (entries proven enough for CLAUDE.md), stale references, consolidation opportunities, conflicts with existing CLAUDE.md rules, and reports health metrics (capacity, freshness, organization). Spawned by `/si:review`.
skill-extractor
/skill-extractor
Transforms a proven pattern or debugging solution into a standalone, portable skill package. Generates `SKILL.md` with proper frontmatter, reference docs, and examples that work in any project (no hardcoded paths or project-specific values). Spawned by `/si:extract` when a recurring solution should become reusable.
Skills3
auto-improve
/auto-improve
Use to autonomously improve skills, memory files, agents, documentation, or conversation-derived workflows based on recent evidence. Runs eval-loop style review and mutation, keeps score and change logs, preserves reusable lessons, and chooses the highest-leverage improvement targets without waiting for explicit user instruction.
memory-management
/memory-management
Design, audit, and improve agent memory systems across Claude Code's three memory layers (CLAUDE.md, auto-memory, rules/). Use for: reviewing what Claude has learned, promoting proven patterns to enforced rules, extracting recurring solutions into skills, checking memory health and capacity, capturing a pattern explicitly, or wiring up the error-capture hook.
second-brain
/second-brain
Build and maintain a durable, storage-agnostic personal or team knowledge system from notes, meetings, docs, chats, research, code repos, datasets, images, and loose ideas. Use when the user wants to capture raw material, incrementally compile it into a maintained LLM wiki, preserve provenance, consolidate duplicates, run recurring health checks, challenge ideas against prior knowledge, answer complex questions against the maintained wiki, generate reports or rendered artifacts from an evolving knowledge base, or turn recent knowledge updates into recurring publishable outputs without depending on a specific tool such as Obsidian.
The plugin manifest points to a different repository than the source indexed by ClaudePluginHub.
Stats
Version2.0.0
LanguagePython
Stars0
Copy clicks1
MaintenanceGood
LicenseMIT
Last Commit
Added
Available In
Safety Signals
Caution
Executes bash commands
Hook triggers when Bash tool is used
Uses power tools
Uses Bash, Write, or Edit tools
Tags
Categories
