Skip to content

Meme-Theory/meme-engine

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

32 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Meme Engine

A Claude Code plugin marketplace by Meme-Engine, Inc.

Install

/install-plugin Meme-Theory/meme-engine

research-clab

Scaffold a complete multi-agent research collaboration in any domain. One command generates the full project architecture: agents, skills, behavioral rules, knowledge tooling, session infrastructure, and a Session 0 plan.

/new-research-project

What it builds

your-project/
  .claude/
    agents/          3 infrastructure + N domain specialists
    skills/          11 skills (team launch, synthesis, indexing, ...)
    rules/           6 behavioral rules (epistemic discipline, team conduct, ...)
    agent-memory/    Per-agent persistent memory
  researchers/       Domain reference corpora (14 papers per agent)
  sessions/          Session plans, templates (Formats A-I), framework docs
  tools/             Knowledge schema + index (librarian-maintained)
  CLAUDE.md          Project constitution
  agents.md          Agent & skill registry

How it works

  1. /new-research-project collects your domain and research question
  2. Dispatches coordinator and librarian agents in parallel to scaffold infrastructure and knowledge system
  3. Presents 10 agent archetypes — you pick a team (Skeptic + Calculator mandatory, Workhorses/Dreamers/etc. optional)
  4. You assign persona overlays (real researchers whose intellectual DNA shapes each agent)
  5. /new-researcher stamps each agent definition and deploys scout to fetch papers
  6. Session 0 prompt is ready — run /clab-team to launch your first research session

Agent archetypes

Archetype Cognitive style
Skeptic "Prove it." Demands methodology, controls, pre-registration. Owns confidence assessment.
Calculator "Stop debating. Run it." Produces artifacts, code, data. Resolves debates by computing.
Workhorse "I'll work through this properly." Deep domain expertise, shows every step.
Principalist "What MUST be true?" Structural constraints, thought experiments.
Dreamer "What if this is actually THAT?" Cross-domain pattern finder.
Boundary Guard "Here are the limits." Hard constraints, impossibility results.
Observer "What does the data say?" Data-driven analysis, error bars.
Bridge "The original source says..." Source fidelity, preserves external work.

Skills installed in generated projects

Skill What it does
/clab-team Launch a coordinated multi-agent session from a prompt file
/clab-synthesis Generate synthesis/fusion documents (solo or team mode)
/clab-plan Generate session plans and prompts from a topic
/clab-review Multi-agent collaborative document review
/new-researcher Create a new domain agent with web-fetched papers
/weave Query and maintain the knowledge index
/indexing Build structured index for a researcher folder
/shortterm Collapse and optimize agent memory files
/document-prep Format-aware document toolkit
/team-blast Direct-write broadcast to team agent inboxes
/redact Remove agent-memory references to a session or identifier

Session formats

Nine predefined formats with prescribed team sizes and workflows:

Format Type Team
A First Contact Review 2-3 agents
B Adversarial Debate 2-3 agents
C Collaborative Deep Dive 2-3 agents
D Workshop 2-3 agents per round
E Investigation Arc 2-3 agents
F Decisive Computation 2-3 agents
G Mass Parallel Assessment 3 agents
H Decision Gate 2-3 agents
I Formalization 2-3 agents

Knowledge system

Every generated project includes a schema-driven knowledge graph:

  • tools/knowledge-schema.yaml — Entity types (universal + domain-specific)
  • tools/knowledge-index.json — Structured index (librarian is sole writer)
  • /weave skill for querying and maintaining the index

The librarian agent indexes session outputs, tracks constraints, gates, proven results, data provenance, and cross-references — all without Python.


Also included

graceful-handoff/grace generates a detailed conversation summary so a new session can pick up where you left off.

skeptic-validator/validate demands concrete evidence (test results, logs, benchmarks) that code changes are correct before accepting them.

License

MIT

About

Windows Compatible ralph-loop skill/hook using Python in lieu of bash

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors