What problem does it solve? Teams adopting AI coding agents on existing (brownfield) projects often don't know whether their current stack helps or hinders agent workflows. This Skill evaluates the detected stack against four quality gates—typed, convention-based, popular in training data, and well-documented—and produces concrete compensation strategies instead of recommending a rewrite. ## Core Features & Use Cases - File-driven stack detection: Reads project markers (package.json, pyproject.toml, Cargo.toml, go.mod, Gemfile, composer.json, *.csproj, pubspec.yaml) plus CI, deployment, and instruction-file signals to identify language, framework, build tool, test runner, and package manager. - Evidence-based gate scoring: Scores each component against the four agent-friendly criteria, citing the specific file or config that proves each pass or fail, with per-language-family assessment for training-data popularity. - Compensation strategies: For every failed gate, generates ready-to-paste AGENTS.md rule entries (type annotation conventions, folder structure docs, framework idiom examples) so agents can work effectively despite gaps. - Use Case: A developer with an existing Express + JavaScript API runs the assessment, learns the stack fails the typed and convention-based gates, and receives paste-ready AGENTS.md rules plus a verdict of ready-with-compensation written to context/foundation/stack-assessment.md. ## Quick Start Ask the AI to assess my stack for agent-friendliness in the current project directory.
