Sandy-zippy Diagnose
Reports are not plans. This skill turns the audit and visibility reports into a ranked, do-this-first plan a non-SEO can execute.
Inputs
seo-reports/audit-<date>.mdandseo-reports/visibility-<date>.md(run those skills first if missing).- DataForSEO MCP for question mining.
Steps
- Merge findings: pull every P0/P1 from the audit and every "why you're skipped" reason from the visibility report into one list. Dedupe (a missing schema finding and a "no corroboration" skip-reason may share one fix).
- Mine the unanswered questions (DataForSEO): for the site's top 3 topics, pull People Also Ask questions and related/long-tail queries with question modifiers (how, best, vs, worth it, near me). Cross-check each against the site: does ANY page answer it? Collect the misses — these are the questions people ask AI where you are silent.
- Rank by effort vs impact: score each candidate fix Impact (1-5: traffic/citation potential, from real impressions and visibility gaps) and Effort (1-5: one-line change → new content). Sort by Impact/Effort descending. Take the top 15.
- Write the plan: each fix gets: what, why (evidence from the reports), how (concrete: "add Organization JSON-LD with sameAs links to LinkedIn + Google Business profile"), and effort (minutes/hours/days).
Output
DIAGNOSIS: <domain> — <date>
YOUR 15 FASTEST FIXES (ranked)
1. <fix> — impact <n>/5, effort <n>/5 — why: <evidence> — how: <one line>
...
QUESTIONS PEOPLE ASK AI THAT YOU NEVER ANSWER
1. "<question>" — <monthly volume or PAA source> — suggested page/section: <where it should live>
...
HAND-OFF
Fixes 1-N are content/schema changes → run the fix skill.
Fixes N+1... need human/dev decisions: <list>
Save to seo-reports/diagnosis-<date>.md. The fix skill consumes it.
Rules
- Never pad to 15. If only 11 fixes are real, ship 11 and say so.
- Impact scores must trace to a number in the upstream reports (impressions, position, citation gap), not gut feel.