Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

SlothSpec: I taught my laptop to redline requirements, then pointed it at a real NASA spec

Description and use

SlothSpec automates requirement review, reducing defects and speeding up the process while preserving all critical issues.

Problem

Reviewing large requirements documents is tedious because many requirements contain vague language, missing units, or other common defects that reviewers must flag manually.

Outcome

SlothSpec automates requirement redlining, reducing reviewer effort and improving consistency while keeping all critical defects intact.

Workflow

  • Drop a spec file (.docx, .pdf, .txt, .md) into the input folder
  • Run SlothSpec locally on a laptop with 16 GB RAM using Ollama and Gemma 4
  • Pattern checks flag obvious defects (vague words, TBDs, numbers without units)
  • Each requirement is sent to Gemma 4 for grading (good/needs work/rewrite) and rewrite suggestion
  • Redline report is generated in the reviewed/ folder
  • Optional: use watch mode, cache, or CSV export for integration into comment matrices

Evidence

The PDF reader fell apart first. v0.3 pulled 451 "requirements" out of it, including page footers glued onto sentences, the table of contents, the revision history, and the name of a referenced standard

Source

Then the model flagged 94% of it. 312 of 333 requirements, in a published, approved NASA standard. Nobody reads 312 comments.

Source

The result was 333 clean shall requirements, every one with its real ID.

Source

The real catches are the kind of thing I'd actually put in a comment matrix: 4.2.3.4j: hazards "mitigated to an acceptable level." Acceptable to whom? →

Source

  • Record date: 2026-10-23T00:26:11.926Z