The /dr-edit Command — Editorial Review Before Publication
How /dr-edit runs three distinct editorial phases — fact verification, AI pattern removal, and structural polish — before any piece of content goes live.
A draft is not finished content. /dr-edit is the stage between draft and publication: it activates the Editor agent and runs three sequential phases of review, each targeting a different class of problem.
The command takes a file path as its argument. If the content came from /dr-write, the editorial notes from that stage carry forward. Before any changes are made, a timestamped backup of the original is created next to the file.
Phase 1: fact verification
Every verifiable claim in the document is extracted and categorized by type — statistics, dates, product names, technical specifications, attributed quotes. Each claim receives an importance rating, which determines how many independent sources it needs. A central thesis claim marked as critical requires three or more sources; a background fact marked as medium needs one.
Verification follows a source hierarchy: official product pages and documentation rank above press releases, which rank above news outlets. Social media posts are not treated as authoritative sources. Each claim gets a verdict: ACCURATE, INACCURATE, OUTDATED, MISLEADING, UNVERIFIABLE, or NEEDS_CONTEXT. Inaccurate and outdated claims are corrected; the correction and its source are logged.
Phase 2: AI pattern removal
The Editor agent scans for writing patterns that mark text as AI-generated. This covers banned vocabulary — words like "delve," "tapestry," "leverage," "robust," "seamless" — as well as structural tells such as consecutive sentences with the same subject or opening word, and communication artifacts like sycophantic openers and knowledge-cutoff disclaimers.
Russian-language content gets its own scan: textbook tone, restated ideas, generic phrases that add length without adding meaning. Changes are applied in three passes — vocabulary first, then structure and rhythm, then a final anti-AI audit — to avoid fixes that introduce new problems.
Phase 3: editorial polish
The third phase checks consistency across the entire document: terminology used uniformly, tone matching the target register, formatting applied correctly. Structural issues — argument flow, section imbalance, weak transitions — are flagged and corrected. All links and citations are verified.
The final naturalness check asks one question about every paragraph: does this read as if a person wrote it? Paragraphs that don't pass get another revision.
Output and handoff
The editorial report groups all changes by category: factual corrections with sources cited, AI pattern fixes with counts, structural improvements, and any changes that altered meaning and need author attention. Changes apply to the original file only after review. The backup stays in place.
After the editorial pass, /dr-publish takes the approved content and handles platform-specific formatting. For targeted micro-checks without the full pipeline, /factcheck covers only fact verification and /humanize covers only AI pattern removal.
Read what Datarim is for context on the pipeline, or see the /dr-write command for the drafting stage that precedes this one.