lint

name	lint
description	Run all configured linters across the full project, including whole-project-only linters. Use when user says "/lint", "lint project", or "run linters".
tools	Bash, Read

Lint

Run configured linters across the full project. Includes whole-project-only linters (brakeman, clippy) that don't run in the PostToolUse hook.

Usage

• /lint — Run all enabled linters on the full project
• /lint ruby — Run only Ruby linters
• /lint reek — Run only the reek linter
• /lint semgrep — Run only Semgrep (tool linter)
• /lint --fix — Run autofix on all files first, then lint

Arguments after /lint are parsed as:

• A language name (ruby, javascript, python, go, rust, markdown, html, shell) → filter to that language
• A linter name (rubocop, reek, eslint, ruff, etc.) → filter to that specific linter
• A tool linter name (semgrep) → filter to that tool linter
• --fix → run autofix commands before linting

Step 1: Load Config

PROJECT_DIR="${CLAUDE_PROJECT_DIR:-$(pwd)}"
PROJECT_HASH=$(echo "$PROJECT_DIR" | tr '/' '-' | sed 's/^-//')
CONFIG_FILE="$HOME/.claude/code-lint/$PROJECT_HASH/config.json"

If no config exists, tell the user to run /setup-lint first.

Read and parse the config JSON.

Step 2: Determine Scope

Parse arguments to determine what to run:

• No args → all enabled languages, all enabled linters
• Language arg → only that language's linters
• Linter arg → only that specific linter
• --fix → enable autofix mode

Step 3: Run Linters

For each enabled linter in scope:

Single-file linters

Find all matching files (using file_patterns), excluding paths that match exclude_patterns. Run the linter on all matching files at once (most linters accept multiple files or directories).

cd "$PROJECT_DIR"

# If --fix mode, run autofix first
bundle exec rubocop -A --fail-level=error .

# Then run the lint check
bundle exec rubocop .

Whole-project-only linters

Run these against the project root:

cd "$PROJECT_DIR"
bundle exec brakeman -q
cargo clippy -- -D warnings

Tool linters (e.g. Semgrep)

After running per-language linters, check the tool_linters section of the config. For each enabled tool linter:

If the argument to /lint matches a tool linter name (e.g. /lint semgrep), run only that tool linter.

For Semgrep, build the command with configured rulesets:

cd "$PROJECT_DIR"

# Build --config flags from rulesets array
# e.g. rulesets: ["p/default", "p/security-audit"] → --config p/default --config p/security-audit
semgrep scan --quiet --config p/default --config p/security-audit .

# With --fix mode:
semgrep scan --autofix --quiet --config p/default .

Respect exclude_patterns — pass each value as a --exclude flag (these are path prefixes/globs, not regex):

semgrep scan --quiet --config p/default --exclude "node_modules/" --exclude "vendor/" .

Collect results

For each linter, capture:

• Exit code (0 = pass, non-zero = issues found)
• stdout/stderr output
• File count checked
• Issue count (parse from output where possible)

Step 4: Present Results

Group results by language and linter:

## Lint Results

**Project:** /path/to/project

### Ruby (42 files)

**rubocop** — 3 issues
  app/models/user.rb:15:5 Style/GuardClause: Use guard clause
  app/models/user.rb:28:3 Metrics/MethodLength: Method too long (15/10)
  app/services/auth.rb:8:1 Layout/EmptyLines: Extra blank line

**reek** — 1 issue
  app/models/user.rb:10 FeatureEnvy: process refers to config more than self

**brakeman** — No issues

### Shell (5 files)

**shellcheck** — No issues

---

**Summary:** 4 issues found (3 rubocop, 1 reek)

Step 5: Offer Fixes

If issues were found and autofix is available:

1. Ask if the user wants to auto-fix fixable issues
2. Run autofix commands
3. Re-run lint to show remaining (non-auto-fixable) issues
4. For remaining issues, offer to fix them using the reference docs for guidance

Read the relevant reference docs for fix patterns:

• Ruby: ${CLAUDE_PLUGIN_ROOT}/references/ruby-patterns.md
• JS/TS: ${CLAUDE_PLUGIN_ROOT}/references/js-ts-patterns.md
• Python: ${CLAUDE_PLUGIN_ROOT}/references/python-patterns.md
• Go: ${CLAUDE_PLUGIN_ROOT}/references/go-patterns.md
• Rust: ${CLAUDE_PLUGIN_ROOT}/references/rust-patterns.md
• Markdown: ${CLAUDE_PLUGIN_ROOT}/references/markdown-patterns.md
• HTML: ${CLAUDE_PLUGIN_ROOT}/references/html-patterns.md
• Shell: ${CLAUDE_PLUGIN_ROOT}/references/shell-patterns.md
• Semgrep: ${CLAUDE_PLUGIN_ROOT}/references/semgrep-patterns.md

If CLAUDE_PLUGIN_ROOT is not set, find references relative to this SKILL.md (two directories up: ../../references/).

Behavior

1. Always cd to project root before running linters
2. Respect timeout settings from config
3. If a linter times out, report it and continue with others
4. Present results in a readable, grouped format
5. Count total issues and offer summary
6. If no issues found, say so concisely
7. If no config exists, suggest /setup-lint