code-standards-adopter

# Code Standards Adopter Most teams' real conventions live in the code, not in a style guide. This skill reads the codebase, infers the conventions actually in force, and turns them into enforceable artifacts — so newly written code (human or AI) matches what's already there instead of fighting it. The output is descriptive first, prescriptive second: capture what IS, flag inconsistencies, and let the team decide what SHOULD be. ## Process 1. **Sample the codebase.** Don't read everything. Pick 10-20 representative files: the most-recently-changed files (current style), the largest modules (dominant style), and one file per major directory. Note the languages and frameworks in play. 2. **Infer conventions per dimension** (see checklist below). For each, record: the dominant pattern, observed frequency (e.g. "camelCase in 18/20 files"), and exceptions worth flagging. 3. **Detect existing enforcement.** Check for linter/formatter configs (.eslintrc, ruff.toml, .editorconfig, checkstyle, prettier, gofmt assumptions), CI lint steps, and pre-commit hooks. Never generate a config that conflicts with one that exists — propose amendments instead. 4. **Generate the artifacts** the user needs (ask if unclear, default to all three): - **Conventions document** — concise markdown, one section per dimension, with real examples lifted from the codebase (anonymized if sensitive). - **Tool configs** — linter/formatter configuration matching the inferred style, with each non-default ...