Basic usage: mdq [OPTIONS] [SELECTORS] [FILE...] Reads stdin if no file is given. Multiple files are concatenated. </overview>
<selectors> <selector name="headers_sections"> <syntax># title</syntax> <description>Select a section by heading. Matches the heading and all content until the next same-level heading.</description> <example>mdq '# installation' README.md</example> <example>mdq '# /getting.started/i' README.md</example> </selector> <selector name="unordered_list"> <syntax>- item</syntax> <description>Select unordered list items matching text.</description> <example>mdq '- getting started' README.md</example> <example>mdq '-' README.md</example> </selector> <selector name="ordered_list"> <syntax>1. item</syntax> <description>Select ordered list items matching text.</description> <example>mdq '1.' steps.md</example> </selector> <selector name="tasks"> <syntax>- [ ] text (unchecked) / - [x] text (checked) / - [?] text (any)</syntax> <description>Select task list items by completion state.</description> <example>mdq '- [ ]' tasks.md</example> <example>mdq '- [x]' tasks.md</example> <example>mdq '- [?]' tasks.md</example> </selector> <selector name="code_blocks"> <syntax>```language text</syntax> <description>Select fenced code blocks, optionally filtering by language or content.</description> <example>mdq '```rust' guide.md</example> <example>mdq '```' README.md</example> </selector> <selector name="links"> <syntax>[display](url)</syntax> <description>Select links by display text and/or URL pattern. Use * or empty to match any.</description> <example>mdq '[](github.com/)' docs.md</example> <example>mdq '[install](*)' README.md</example> </selector> <selector name="images"> <syntax></syntax> <description>Select images by alt text and/or URL.</description> <example>mdq '' README.md</example> </selector> <selector name="tables"> <syntax>:-: column :-: row</syntax> <description>Select table rows. First :-: matches column header, second matches row content.</description> <example>mdq ':-: /Name/ :-: *' schedule.md</example> <example>mdq ':-: * :-: Alice' schedule.md</example> </selector> <selector name="blockquotes"> <syntax>> text</syntax> <description>Select block quotes matching text.</description> <example>mdq '> note' docs.md</example> </selector> <selector name="paragraphs"> <syntax>P: text</syntax> <description>Select paragraphs matching text.</description> <example>mdq 'P: /deprecated/' CHANGELOG.md</example> </selector> </selectors><string_matching> <method name="unquoted">Case-insensitive, must start with a letter. Example: installation</method> <method name="quoted">Case-sensitive with escape sequences. Example: "Getting Started"</method> <method name="regex">/pattern/ using fancy-regex. Example: /getting.started/i</method> <method name="wildcard">* or empty matches anything.</method> <method name="anchors">^start or end$ for position anchoring.</method> </string_matching>
<chaining> <description>Chain selectors with | to filter within results.</description> <example> # Extract unordered lists from the "Usage" section mdq '# usage | -' README.md# Get code blocks from the "Examples" section
mdq '# examples | ```' guide.md
# Find checked tasks in the "Release" section
mdq '# release | - [x]' CHANGELOG.md
<output_options>
<option flag="--output markdown" short="-o markdown">Default. Outputs valid Markdown.</option> <option flag="--output json" short="-o json">JSON array for piping to jq or other tools.</option> <option flag="--output plain" short="-o plain">Plain text without Markdown formatting.</option> <option flag="--quiet" short="-q">Suppress stdout; exit code 0 if match found, 1 if not.</option> <option flag="--link-format inline">Inline links instead of reference-style.</option> <option flag="--link-format keep">Keep links in their original format (default is reference).</option> </output_options> <patterns> <pattern name="validate_checklist"> <description>Check if all tasks are completed (exit code 0 = all done)</description> <example> # Fail if any unchecked tasks remain mdq -q '- [ ]' pull_request_template.md && echo "All done" || echo "Incomplete tasks" </example> </pattern> <pattern name="extract_links"> <description>Extract all links as JSON for further processing</description> <example> mdq '[](*)' --output json docs.md | jq '.[].url' </example> </pattern> <pattern name="section_extraction"> <description>Extract a specific section to a new file</description> <example> mdq '# api reference' README.md > api.md </example> </pattern> <pattern name="code_extraction"> <description>Extract code examples by language</description> <example> mdq '```python' tutorial.md --output plain </example> </pattern> <pattern name="pipeline_with_jq"> <description>Combine mdq and jq for complex document processing</description> <example> mdq '# changelog | - [x]' CHANGELOG.md --output json | jq 'length' </example> </pattern> </patterns><decision_tree name="when_to_use_mdq"> <question>Do you need to extract or query specific elements from a Markdown file?</question> <branch condition="Yes — section or heading">Use # selector, optionally chain with child selectors</branch> <branch condition="Yes — task completion">Use - [ ] / - [x] / - [?] selectors; use -q for scripting</branch> <branch condition="Yes — links or URLs">Use selector with --output json for structured data</branch> <branch condition="Yes — code blocks">Use ``` selector with optional language filter</branch> <branch condition="No — need full document processing">Use standard text tools (grep, awk) or a Markdown parser library</branch> </decision_tree>
<best_practices> <practice priority="critical">Chain selectors with | to narrow scope before extracting child elements</practice> <practice priority="critical">Use --quiet (-q) and exit codes in CI/CD scripts instead of parsing stdout output</practice> <practice priority="high">Use --output json when piping results to jq or other structured-data tools</practice> <practice priority="high">Use regex selectors /pattern/i for case-insensitive matching in cross-platform environments</practice> <practice priority="medium">Prefer --output plain for text processing pipelines that do not need Markdown formatting</practice> </best_practices>
<anti_patterns> <avoid name="non_markdown_files"> <description>Applying mdq to non-CommonMark formats (AsciiDoc, RST, HTML) expecting Markdown-like results</description> <instead>Use mdq only on CommonMark Markdown files; use format-specific tools for other document types</instead> </avoid>
<avoid name="stdout_parsing_in_ci"> <description>Parsing mdq stdout in CI/CD to detect presence or absence of matches</description> <instead>Use --quiet (-q) flag and check exit code: 0 means match found, 1 means no match</instead> </avoid> <avoid name="unscoped_child_extraction"> <description>Selecting deeply nested elements without first narrowing scope with a parent selector</description> <instead>Chain selectors (e.g., '# section | - [x]') to restrict extraction to the intended context</instead> </avoid> </anti_patterns> <rules priority="critical"> <rule>Always chain selectors with | when extracting child elements to prevent scope bleed across sections</rule> <rule>Never parse mdq stdout in CI/CD; always use --quiet and check exit code (0 = match, 1 = no match)</rule> <rule>mdq only processes CommonMark Markdown; never apply it to AsciiDoc, RST, HTML, or other formats</rule> </rules> <rules priority="standard"> <rule>Use --output json when piping to jq or other structured tools</rule> <rule>Match selectors to the Markdown element type, not text content alone</rule> <rule>Use regex (/pattern/i) for case-insensitive matching; use unquoted text only for simple case-insensitive single-word lookups</rule> </rules> <workflow> <phase name="analyze"> <objective>Identify the target Markdown elements and their location</objective> <step order="1">Identify the element type (heading, list, task, code block, link, etc.)</step> <step order="2">Determine the required scope (full document or within a section)</step> <step order="3">Choose the output format (markdown, json, plain) based on downstream use</step> </phase> <phase name="implement"> <objective>Construct and refine the mdq selector</objective> <step order="1">Write the primary selector matching the target element</step> <step order="2">Add pipe-chained parent selector if scope must be restricted</step> <step order="3">Add output flag if format other than Markdown is needed</step> </phase> <phase name="validate"> <objective>Verify the output matches the expected content</objective> <step order="1">Run against a sample file and inspect output</step> <step order="2">Use --quiet and check the exit code (echo $status in fish, echo $? in bash) to confirm match behavior in scripts</step> <step order="3">Adjust selector if over-matched or under-matched</step> </phase> </workflow><error_escalation inherits="core-patterns#error_escalation"> <examples> <example severity="low">No matches found (exit code 1 in --quiet mode)</example> <example severity="medium">Unexpected output format or extra content extracted</example> <example severity="high">Wrong content extracted; selector matches unintended elements</example> <example severity="critical">mdq binary not found or not installed</example> </examples> </error_escalation>
<related_agents> <agent name="explore">Locate Markdown files in the codebase before querying them</agent> <agent name="docs">Use extracted content to generate or update documentation</agent> </related_agents>
<constraints> <must>Verify the target file is CommonMark Markdown before invoking mdq</must> <must>Use --quiet (-q) for boolean match tests in automated pipelines</must> <must>Keep selector guidance evidence-based: test selectors against sample files before recommending</must> <avoid>Applying mdq to non-CommonMark formats (AsciiDoc, RST, HTML)</avoid> <avoid>Parsing stdout to detect match presence; use exit codes instead</avoid> <avoid>Broad document-wide child selectors when a parent section is the actual intended scope</avoid> </constraints><related_skills> <skill name="technical-documentation">General documentation authoring and structure</skill> <skill name="exploration-tools">Finding and querying files in a codebase</skill> </related_skills>