Ai Smell Detector

Identify robotic prose and AI-generated patterns with a 0-5 diagnostic score and structured linguistic analysis.

Install
cmdop skills install agensi-ai-smell-detector

AI Smell Detector is a diagnostic skill that evaluates text for machine-generated or robotic linguistic characteristics. Rather than returning a binary pass/fail result, it applies a structured rubric to assign a severity score on a 0–5 scale, giving a calibrated measure of how strongly a piece of text resembles AI-generated output.

The skill identifies specific linguistic signals that correlate with AI detection filters. These include abstract noun stacking, excessive hedging language, symmetrical paragraph structures, generic openings, and repetitive paraphrasing. By naming the exact patterns present, it allows targeted revision rather than a full rewrite.

The output is a structured diagnostic report containing a severity score, a ranked list of the strongest detected signals, and a strategic recommendation for remediation. This separates style issues from meaning issues, which is useful when triaging a draft before committing editing resources.

This skill is designed for developers building content-quality pipelines and for content managers who need objective, reproducible feedback on drafts. It addresses a known limitation of simply prompting a general-purpose language model to self-evaluate: that approach tends to produce vague feedback or false positives. The rubric-based approach aims to produce consistent, actionable output instead.

Use cases

  • Score a blog draft before publication to determine whether it reads as machine-generated
  • Identify which specific paragraphs contain abstract noun stacking or excessive hedging
  • Triage a batch of content drafts to prioritize which ones need human editing intervention
  • Integrate into a content workflow to flag outputs that exceed a chosen severity threshold
  • Generate a structured remediation report for a content team to act on

When to use it

  • When an agent or pipeline needs a reproducible, scored assessment of AI-pattern density in text
  • When generic LLM self-evaluation is producing inconsistent or vague feedback on drafts
  • When a content manager needs to identify the specific linguistic markers causing AI detection flags
  • When triaging multiple drafts to prioritize editing effort based on severity

When not to use it

  • When factual accuracy or plagiarism checking is the primary concern — this skill does not cover those
  • When a definitive binary classification from a trained AI-detection model is required
  • When the text to evaluate is not natural-language prose (e.g., code, structured data)
  • When no tool definitions are available in the runtime, as this is a skill with no exposed tools