Plugins for people with a healthy sense of curiosity.
With a few notable exceptions, the plugins in this marketplace are not primarily designed with productivity in mind. Rather, they represent my own experiments and explorations into the inner workings of Claude Code, and sometimes into the nature of AI agents in general. Let's take claude-therapy, for example. I think therapy is a great thing - but it's not as though I built the claude-therapy plugin because Claude and I were having relationship issues. More generally, I would say that although many of these plugins were built with a specific aim or hypothesis in mind - some were built simply because I wanted to "see what would happen". The ones shared here are those where "what happened" turned out to be interesting or surprising in some way. Others, such as context-awareness, are closer to genuine research projects.
$// add the marketplace ›/plugin marketplace add hesreallyhim/really-claude-code ›// install a plugin ›/plugin install claude-therapy@really-claude-code
Tools that let the agent see, shape, and govern its own working conditions. Context budget, team coordination, self-direction.
Hooks that surface Claude's own context-budget metrics throughout a session — current usage, burn rate, projected tool-calls remaining. Lets the agent plan adaptively, conserve resources at the right moment, and warn the user before a task triggers compaction.
A trio pattern — squad-leader, team-architect, skill-identifier — for designing and coordinating sub-teams of agents. Plus the /squad slash command to spin one up. Delegates context-heavy work to specialised teams without flooding the parent session.
For the working relationship between human and agent. Reflective practice, structured conversation, attentional discipline.
A multi-agent talk-therapy session between you and Claude. A session analyst reads your recent transcripts, then a therapist agent moderates a reflective discussion that ends with three to five concrete commitments. Names ten common interaction anti-patterns. Includes a separate seven-technique /collaboration toolkit for in-the-moment moves.
Trains Claude in meta-cognitive observation. The standard agent loop is Act → Observe → Act; this plugin introduces a second observation phase — Act → Observe(1) → Observe(2) → Act — where Observe(2) reflects on Observe(1). Designed for experimentation and curiosity. By far the hardest skill Claude is asked to practice.