What problem does it solve? Sokrates' init command generates a workable _sokrates/config.json, but it cannot know which files are real source versus vendored, generated, or test code, how the system should be decomposed into components, or which cross-cutting concerns matter. Misconfigured scope and decomposition silently corrupt every metric in every report, and malformed regexes fail invisibly because Sokrates treats them as matching nothing. ## Core Features & Use Cases - Config creation and tuning: Sets source scope (srcRoot, extensions, ignore rules), classifies files into main/test/generated/build/other scopes with correct precedence, and designs logical decompositions via folder depth, explicit components, or meta rules. - Preview simulation script: Runs scripts/preview_config.py to apply Sokrates' own scoping rules to the real tree before analysis, reporting excluded files, scope sizes, component LOC shares, concern hits, and linting dead rules and non-compiling regexes. - Field reference: references/config-reference.md documents every config key read from the Sokrates Java source, including gotchas the official docs get wrong (e.g. trendAnalysis and maxFileSize do not exist). - Use Case: A team sets up Sokrates across 30 repositories; the skill writes a shared analysis_conventions.json, runs init -conventionsFile per repo, and verifies one representative repo with the preview script before rollout. ## Quick Start Ask the AI to set up or fix the Sokrates configuration for your repository and preview what the config will include and exclude before running the analysis.
