Tool · Active
Markdown Cleaner
Paste whatever a chatbot put on your clipboard — even after Bear, Notion, Google Docs, or another tool in the round-trip chain has mutated it. The Cleaner normalizes heading hierarchy, bullet markers, list spacing, table format, and special characters into AIX-conformant output, and flags structural ghosts that copy-chain damage leaves behind.
Tip: press ⌘ + Enter to clean.
What the Cleaner does
13 transformations, applied in order.
- Strip invisible characters: U+2011 non-breaking hyphens, U+2800 Braille pattern blanks, U+200B zero-width spaces, U+2060 word joiners, U+00AD soft hyphens, U+FEFF BOMs.
- Remove backslash-escaped periods from numbered list markers (
1\.→1.). - Convert bold-on-its-own-line pseudo-headings into real H2 or H3 headings (
**1. LangGraph**→## 1. LangGraph,**Key Features**→### Key features). - Normalize subsection labels:
### Key Features:→### Key features(sentence case, no colon). - Replace asterisk and plus bullets with ASCII hyphen-minus.
- Remove blank lines between adjacent same-level list items.
- Normalize nested-list indentation — tabs and 4-space → 2-space.
- Replace
***,___, and long dash runs with---. - Reformat tables: ensure alignment row, single-space cell padding, no decorative bold around canonical values, convert glyph cells (✓, ✗, ●, 🟢, 🔴, 🟡) to
Yes,No,Partial. - Normalize emphasis:
_italic_→*italic*and__bold__→**bold**(skipping code blocks, inline code, and identifiers). - Remove unnecessary backslash escapes (
\,\!\?etc.). - Collapse 3+ consecutive blank lines into one blank.
- Trim trailing whitespace from every line.
Each transformation is fully reversible by inspection. The Cleaner does not lose content — it only normalizes form. Round-trip detection runs alongside as a heuristic warning: if the cleaned markdown references “a comparison table” or “the following code” but no such structure exists in the input, the Cleaner flags it as possible copy-loss from the chatbot interface.
Deferred to v0.2: heading-hierarchy auto-repair, citation format normalization, summary-section wrapping, indented-to-fenced code block conversion. These require more semantic judgment than is safe to automate in v0.1.