Scanning Commands

Rulebound provides several commands for scanning code, plans, and diffs against your rules.

rulebound find-rules

Find and filter rules by task, category, tags, or stack.

rulebound find-rules [options]

Flag	Description
`-t, --task <text>`	Describe the task to find relevant rules
`--title <title>`	Search by title
`-c, --category <category>`	Filter by category
`--tags <tags>`	Filter by tags (comma-separated)
`--stack <stack>`	Filter by tech stack (comma-separated)
`-f, --format <format>`	Output: `table` (default), `json`, `inject`
`-d, --dir <path>`	Path to rules directory

Examples:

# Find rules relevant to an auth task
rulebound find-rules --task "add JWT authentication"

# Filter by category and tags
rulebound find-rules --category security --tags "auth,api"

# Output in inject format (ready to paste into agent context)
rulebound find-rules --format inject

# JSON output
rulebound find-rules --format json

The inject format outputs rules with [MUST]/[SHOULD]/[MAY] prefixes, ready to paste into an AI agent's context window.

rulebound validate

Validate an implementation plan against matched rules.

rulebound validate [options]

Flag	Description
`-p, --plan <text>`	Plan text to validate
`--file <path>`	Path to a plan file
`-f, --format <format>`	Output: `pretty` (default), `json`
`-d, --dir <path>`	Path to rules directory
`--llm`	Use LLM for deep validation (requires AI SDK)

Examples:

# Validate inline plan text
rulebound validate --plan "Add a REST endpoint for user deletion"

# Validate from a file
rulebound validate --file implementation-plan.md

# With LLM-powered deep validation
rulebound validate --plan "Refactor auth module" --llm

# JSON output for scripting
rulebound validate --file plan.md --format json

The validation report shows each rule with its status: PASS, VIOLATED, or NOT_COVERED. Exit code 1 if any MUST rule is violated.

rulebound diff

Validate git diff against rules before merge.

rulebound diff [options]

Flag	Description
`--ref <ref>`	Git ref to diff against (default: HEAD)
`-f, --format <format>`	Output: `pretty` (default), `json`
`-d, --dir <path>`	Path to rules directory
`--llm`	Use LLM for deep validation

Examples:

# Diff against HEAD (uncommitted changes)
rulebound diff

# Diff against a specific branch
rulebound diff --ref main

# JSON output
rulebound diff --format json

rulebound check-code

Analyze a source file with AST-based anti-pattern detection using tree-sitter.

rulebound check-code [options]

Flag	Description
`--file <path>`	Path to source file (required)
`-l, --language <lang>`	Language override (auto-detected from extension)
`-q, --queries <ids>`	Comma-separated builtin query IDs to run

Supported languages: typescript, javascript, python, java, go, rust, c_sharp, cpp, ruby, bash.

Examples:

# Analyze a TypeScript file
rulebound check-code --file src/auth.ts

# Specify language manually
rulebound check-code --file config --language typescript

# Run specific queries only
rulebound check-code --file app.py --queries "empty-catch,console-log"

rulebound score

Calculate rule quality score and generate a README badge.

rulebound score [options]

Flag	Description
`-d, --dir <path>`	Path to rules directory
`--no-badge`	Skip badge generation
`-o, --output <path>`	Save badge markdown to a file

Examples:

# Score your rules
rulebound score

# Save badge to a file
rulebound score --output badge.md

Rules are scored on three dimensions: atomicity (0-5), completeness (0-5), and clarity (0-5). The overall score is a percentage out of 100.