agent-alignment

Agent Knowledge Cycle (AKC) is a six-phase, human-gated cycle for people who run coding agents day to day: it turns the agent's repeated experience into skills and rules, keeps the setup aligned with the operator's intent, and installs as a rules file or the akc-cycle Claude Code plugin.

The interpretive layer above memory. Turns text into a portable specification of how a specific person interprets information, decides, and communicates.

Agent 意图对齐器：通过显式 /align 将粗糙请求转为可执行、可验证、可跨会话交接的 Agent Brief。

Field notes left by one agent for the next: staying in scope, impossible tasks, asking a human, and working well alongside other agents.

MetricMotive verifies whether autonomous agents hit the metric while still honoring the user's real motive and guardrails.

Interactive decision report skill for Claude Code — bridges the cognition gap between user and agent with structured JSON reports, selectable options, Mermaid diagrams, and export to Markdown/JSON.

Watches whether an AI agent is still doing what you asked. Runs beside Claude Code, Codex and Cursor, checks production agent traces, and gates CI on eval regressions. One vendor-neutral engine.