cyberneticlibrary

AI team memory and specialist agents

rcodeskillsetup L20
hanzlahabib/rihal-code
What it does

Maintain persistent project memory, orchestrate multi-phase sprints, and track architectural decisions

Best for

Multi-week solo or small-team projects where session-to-session coherence and onboarding context are critical.

Inputs
  • · Project context
  • · Decision records (ADR)
  • · Sprint roadmaps
Outputs
  • · .rcode/memory/ markdown files
  • · SPRINT.md task lists
  • · Agent-readable context at session start
Requires
  • · Claude Code
  • · Cursor
  • · Gemini
  • · VS Code
  • · git
Preconditions

Project initialized with /rcode-new-project; .rcode/ folder exists; decisions.md live in repo

Failure modes
  • · Agent re-explains architecture each session → context not persisted
  • · Late decision shifts with no audit trail → team lost on intent
  • · Specialist review agents lack council state → contradictions undetected
Trust signals
  • · 457 automated tests, 45 agents, 116 commands—actively dogfooded weekly
  • · Decision lives in .rcode/memory/, agent reads at startup (~5K tokens auto-loaded)
  • · Specialist council mode (/rcode-council) runs 5 agents in parallel with round-2 debate