CodeVetter · Repo Unpack · Contributor handoff
typescript-eslint/typescript-eslint
Tests and CI were inspected, not executed, and all described flows are source traces. Future agents should begin with the owning package and preserve public parser, plugin, and generated-config contracts. Test and build commands below describe repository workflows for a separately authorized implementation task; they do not authorize execution during this read-only investigation. This bounded inventory leaves additional packages, rules, fixtures, and browser behavior for task-specific inspection.
- Read the owning manifest, root scripts, and local-development guide first; the guide directs contributors to the packageManager field and explains package build dependencies.
- For authorized validation, start with the owning package’s test command, then relevant lint/typecheck checks; packed integration tests are a separate root script, and Vitest is configured without watch mode.
- Installation has substantial side effects: postinstall changes to workspaceRoot, installs Husky hooks, and normally runs clean, build, and ast-spec type checking unless skip conditions apply.
- Treat generated presets as generator outputs: their headers prohibit hand editing, and rule-development documentation identifies generate-configs and rule-doc snapshot regeneration as follow-up workflows.
- Preserve plugin identity across package imports and compatibility boundaries; parser/plugin manifests support ESLint 8.57/9/10 and TypeScript below 6.1, while parser startup explicitly rejects TypeScript major versions at least 7.
- When changing parser behavior, preserve forced ESLint tokens/comments/locations and node maps; legacy single-run repeated parsing falls back to an isolated Program, whereas projectService bypasses that fallback.
- Repository skill guidance emphasizes comment-preserving fixers, source-level type fixes, shared utilities, precise static rule tests, and deferred checker calls; visitor keys must retain source order rather than alphabetical order.
- The AI contribution policy requires close human review and sufficient independent understanding, discourages AI-generated issue/comment/PR prose, and allows maintainers to close contributions that do not follow its guidance.