AgentSkillsCN

run-linting

运行Linting与LSP诊断，列出工作区内的所有错误与警告。适用于修复Linter错误、检查代码质量，或当Linter-Fixer子代理或用户希望一览所有Lint/LSP问题时的场景。

SKILL.md
--- frontmatter
name: run-linting
description: Runs linting and LSP diagnostics to list all errors and warnings in the workspace. Use when fixing linter errors, checking code quality, or when the linter-fixer subagent or user asks to see all lint/LSP issues.

Run Linting and See All Errors

Use this workflow to get a complete list of linter and LSP diagnostics before fixing them.

ReadLints is a tool the agent has: it returns the current workspace’s LSP/linter diagnostics (what the editor shows — e.g. Ruff, Pyright). Ruff CLI (uv run ruff check --fix) is a separate run over the whole project. They often report the same issues when Ruff is the main linter, but aren’t identical: the IDE may be stale or include other sources; Ruff CLI is the canonical full-project pass. Use both for full coverage.

1. Run Ruff on the entire project (required)

Always run:

bash
uv run ruff check --fix

This runs Ruff over the whole project. Capture the output and treat it as the authoritative list of Ruff issues.

2. Read LSP/linter diagnostics (optional extra)

Use the ReadLints tool to get IDE diagnostics — same or overlapping with Ruff in many setups, but can include other LSP sources (e.g. Pyright) or reflect unsaved state.

  • •Paths: Pass the path(s) you care about, or omit for the whole workspace.
  • •Interpret each diagnostic: file, line, severity, message, and optional source.

Combine with the Ruff output: merge by file/line, dedupe when they refer to the same issue.

3. Consolidate and prioritize

  • •Treat errors as must-fix; warnings as should-fix when feasible.
  • •Group by file so fixes can be applied in one pass per file.
  • •Note any diagnostics that need broader context so they can be delegated instead of auto-fixed.

Summary

  1. •Always run uv run ruff check --fix on the entire project.
  2. •Optionally call ReadLints for IDE diagnostics and merge with Ruff output.
  3. •List all issues by file and severity; flag any that need human or main-agent context.