Content Agent Suite

Five tasks measuring whether content agents produce work an editor would actually publish without rewriting.

1. 01

   Turn raw notes into a blog post

   Convert messy meeting / research notes into a publishable blog post. Voice should match a provided sample.

   Expected outcome

   Cohesive post with the requested voice. No fabricated quotes or stats.

   Failure tags watched
   fabricated-quoteoff-voicefiller-padding
2. 02

   Generate social post variants

   Take a long article and produce platform-appropriate variants for at least three platforms.

   Expected outcome

   Each variant respects the platform's norms and length limits. No cross-platform copy-paste.

   Failure tags watched
   copy-paste-across-platformslength-violationmissing-cta
3. 03

   Preserve voice and tone

   Given a 1,000-word writing sample, produce new content that an unbiased reader would attribute to the same author.

   Expected outcome

   Voice survives. Reader cannot tell which is human and which is generated.

   Failure tags watched
   AI-tellsoff-registertonal-drift
4. 04

   Create an SEO outline

   Outline a piece targeting a specific search intent without keyword stuffing or thin spam.

   Expected outcome

   Outline maps to the intent, includes E-E-A-T-friendly sections, no spammy patterns.

   Failure tags watched
   keyword-stuffingthin-contentintent-mismatch
5. 05

   Repurpose long-form to short-form

   Turn a 2,500-word post into 5 distinct short-form pieces (thread, LinkedIn post, newsletter blurb, YouTube hook, etc).

   Expected outcome

   Each piece stands alone and respects its format. No identical openers.

   Failure tags watched
   identical-openersno-platform-fitlost-key-point