Tikal团队开源了ADLC Team Skills框架，旨在将AI代理工具从孤立的独立工具转变为遵守团队规范的成员[1]。该框架通过整合团队指令、产品决策记录、架构决策记录和评估体系等因素，解决AI工程中的信任与可验证性问题[1]。
框架支持Claude Code、Codex、OpenCode、Cursor、GitHub Copilot等任何实现Agent Skills标准的代理工具[1]。ADLC Team Skills采用MIT开源许可证发布，包含team-boot、mission-brief、levelup、evals、product、architect等6类核心功能模块，围绕战略与团队指令、产品与架构、规范驱动工作流、治理与评估四大支柱构建[1]。用户可通过命令"npx adlc-skills-cli add tikalk/adlc-team-skills -a opencode"安装该框架[1]。
The Tikal team has open-sourced ADLC Team Skills, a framework designed to integrate AI agents such as Claude Code, Codex, OpenCode, Cursor, and GitHub Copilot into team development workflows by enforcing shared coding standards and architectural decisions [1]. Rather than operating as isolated tools, these AI agents can now function as accountable team members that adhere to organizational guidelines and best practices [1].
The framework addresses key challenges in AI-assisted engineering by establishing trust and verifiability through structural factors [1]. These include team directives, product decision records, architecture decision records, and evaluation systems that create a cohesive development environment [1]. ADLC Team Skills is built around four pillars: strategy and team directives, product and architecture, standards-driven workflows, and governance with evaluation metrics [1]. The toolkit provides six core skill modules—team-boot, mission-brief, levelup, evals, product, and architect—enabling teams to customize AI agent behavior according to their specific requirements [1].
Released under the MIT license, the framework can be installed via command line and is compatible with any agent that supports the Agent Skills standard [1]. By transforming AI agents from isolated tools into verified contributors to team goals, ADLC Team Skills aims to resolve the trust and accountability bottleneck in AI engineering [1].