council
๐ฏSkillfrom boshu2/agentops
|AgentOps โ a toolbox of operational primitives for coding agents that preserves learnings between sessions, supports parallel swarm execution, and builds validation pipelines.
Overview
Council is a skill from the AgentOps repository, a toolbox of composable primitives that solve the problem of coding agents forgetting everything between sessions. AgentOps applies DevOps' Three Ways (Flow, Feedback, Learning) to the agent loop, enabling knowledge accumulation across sessions with automatic learning extraction and injection.
Key Features
- Persistent Learning - Every session extracts learnings into an append-only ledger, scores them by freshness, and re-injects the best ones at the next session start, so later sessions benefit from earlier discoveries
- Council Validation - Independent judges review code and decisions before shipping, catching issues through a structured validation gate system
- Swarm Parallelism - Parallelize any skill with swarm mode, running dependency-ordered waves or full research-plan-implement pipelines simultaneously
- Three Ways Framework - Organized into Flow (
/research,/plan,/crank,/swarm,/rpi), Feedback (/council,/vibe,/pre-mortem, hooks), and Learning (.agents/,aoCLI,/retro,/knowledge) - Automatic Decay - Stale insights decay automatically over time, ensuring that only relevant, high-confidence learnings are injected into new sessions
Who is this for?
This skill is designed for developers running multiple AI coding sessions who want their agent to accumulate knowledge and maintain context across sessions rather than starting fresh each time. It is particularly valuable for teams managing complex projects where decisions, patterns, and learnings from past sessions should inform future work.
Same repository
boshu2/agentops(35 items)
Installation
npx vibeindex add boshu2/agentops --skill councilnpx skills add boshu2/agentops --skill council~/.claude/skills/council/SKILL.mdSKILL.md
More from this repository10
Orchestrates multiple Codex agents for parallel task execution, with the lead agent preventing file conflicts before spawning via merge/wave strategies. Auto-selects the runtime backend (Codex sub-agents, Codex CLI, OpenCode subagents, or swarm fallback) based on availability.
AgentOps โ a toolbox of operational primitives for coding agents that preserves learnings between sessions, supports parallel swarm execution, and builds validation pipelines.
AgentOps automates context persistence between coding agent sessions by extracting learnings, quality-gating them, and injecting them into the next session. Applies DevOps' three ways (flow, feedback, continual learning) to the agent loop with composable primitives for research, planning, building, and review.
AgentOps โ a toolbox of operational primitives for coding agents that preserves learnings between sessions, supports parallel swarm execution, and builds validation pipelines.
AgentOps โ a toolbox of operational primitives for coding agents that preserves learnings between sessions, supports parallel swarm execution, and builds validation pipelines.
Streamlines AI agent development with observability, performance tracking, and debugging tools for complex multi-agent workflows and LLM interactions.
A full discovery phase orchestrator skill that coordinates brainstorming, design validation, codebase research, epic planning, and pre-mortem analysis through strict delegation to sub-skills. Produces execution packets with structured plans for implementing features or capabilities.
An SDLC control plane for agentic software development that adds phase gates, a persistent learning corpus, and multi-agent council reviews on top of existing coding tools like Claude Code, Codex, and Cursor. Captures decisions, planning rules, and learnings in a local .agents/ directory so context survives across sessions.
An SDLC control plane for agentic software development that adds phase gating, a persistent learning corpus, and decision tracking on top of coding agents like Claude Code and Codex, with all state stored in .agents/ alongside your code.
Part of AgentOps, an SDLC control plane for agentic software development that adds phase gating, decision recording, and a persistent learning corpus on top of coding agents like Claude Code, Codex, and Cursor. Stores learnings, patterns, planning rules, and cited decisions in a local .agents/ directory that survives across sessions.