What it does
AIWG gives an AI assistant reusable project context and specialist workflows. Its deployment command copies Markdown and YAML source files into the locations each supported provider reads, so one set of sources works across 18 named provider integrations, plus a generic adapter for other harnesses. It installs from npm as the aiwg package, and aiwg use all with a provider name deploys it into a project. The building blocks are agents (specialist roles such as Security Auditor or Test Architect), skills, commands, rules, behaviors and templates. Workflows save requirements, architecture decisions and other outputs in a .aiwg directory so later sessions can build on them, and a review can pass through several specialist agents before a human approval step. The page also describes agent loops that verify results and adapt after failures, traceability links between documents, code and tests, phase gates, and voice profiles for writing style. You describe the outcome you want in your AI tool and the assistant finds and follows the matching AIWG procedure. The site states that AIWG does not replace your provider subscription or the need to review generated work.
Capabilities
- One set of agents, skills and rules deployed to 18 named AI providers
- Persistent .aiwg memory for requirements, decisions and test strategies
- Multi-agent review workflows with a synthesis step and human approval
- Agent loops that run verification and adapt after failures
- Phase gates and traceability between documents, code and tests
- An aiwg doctor command for checking an install
When to use it
- Reusing the same agent roles and rules across several AI coding tools
- Keeping project decisions and requirements available between AI sessions
- Running structured reviews of a design or code from several specialist perspectives
