Summary

Fixes #114.

parseTags() in src/utils.ts now emits a parse-error (level: critical) when a tag has an unterminated string attribute, instead of silently discarding the tag as plain text.

Root Cause

findTagEnd() (utils.ts:26–51) uses a state machine with three states: normal, string, and escape. When it sees " in STATES.normal it transitions to STATES.string and stays there until a closing " is found. If the closing quote is missing (e.g. {% quote content="test /%}), the scanner never returns to STATES.normal, so it never matches the %} as a valid tag end and returns null.

In parseTags(), the end == null branch (previously lines 87–91) simply advanced pos past {% and called continue — the entire malformed construct was silently treated as plain text with no diagnostic.

Solution

When findTagEnd returns null, check whether a raw %} exists later in the content (via content.indexOf('%}', ...)):

• %} present but unreachable (unterminated string): push an error token with message "Unterminated string in tag attribute". The parser in parser.ts:131–134 promotes this into a parse-error node with level critical on the AST, making it visible to validators and error reporters.
• No %} at all: {% is not a Markdoc tag (e.g. code examples that happen to contain {%). Preserve the existing skip behaviour so no regression occurs for these cases.

The change is 28 lines in parseTags() and adds no new dependencies.

Testing

• Added parseTags — malformed tags — generates an error token for an unterminated string attribute — reproduces the exact scenario from the issue ({% quote content="test /%}); previously produced zero error tokens, now produces one.
• Added parseTags — malformed tags — leaves as text when closing delimiter is entirely absent — asserts that {% with no %} anywhere (e.g. code blocks) is still left as plain text, preventing regression.
• Added parseTags — malformed tags — still parses valid tags after a malformed one — asserts that recovery works and the rest of the content is parsed normally.
• All 262 existing specs continue to pass.

Run with:

npm test

Checklist

• Fixes the root cause (not just the symptom)
• New tests cover the exact failing scenario from the issue
• All existing tests pass (262 specs, 0 failures)
• No unrelated changes
• Code style matches project conventions
• Read CONTRIBUTING.md and followed its requirements