self-improving plugins

by anshss

Self-improving multi-agent harness that grinds your ticket backlog across every repo in your product — guarded auto-merge on green CI, and every resolved ticket writes a lesson into your git-tracked CLAUDE.md. Every ticket starts on a cheap tier and escalates once, on measured failure, instead of being predicted expensive up front; deterministic fixes skip the model entirely; and the lessons file evicts rather than accumulates, so always-on context stays a fixed budget.

by Artience

Cognitive orchestration framework with dual-process routing, lifelong learning, expert agent teams, and workflow automation for Claude Code

by akbun

Agent operations skills — set up a project's memory layout once, rebuild working context at session start, and turn each session's lessons into approved edits to the memory structure the codebase already has

by Denys Rotnov

iEvo — self-evolving plugin for AI coding agents (Claude Code, Codex). Universal via the agentskills.io standard. Capture lessons, patch local agents and skills, replay logs on upstream updates. Antivirus-style security audit before install.

by Denys Rotnov

iEvo — self-evolving plugin for AI coding agents (Claude Code, Codex). Universal via the agentskills.io standard. Capture lessons, patch local agents and skills, replay logs on upstream updates. Antivirus-style security audit before install.

by Jan Kolenko

Composable, portable agent skills that carry a ticket from Jira or Confluence Data Center through branch, tests and review to an open pull request.

by Kadihx

Offline $0 decision layer for coding agents: Choice/Score/Noul primitives, a BELKI confidence gatekeeper, ultra-planning, red-teaming, 4-channel research, lossless log compaction and RLVR self-improvement — as an MCP server + CLI + skill.