Source Command Audit Geo

source-command-audit-geo

Use this skill when the user asks to run the migrated source command audit-geo.

Command Template

/audit-geo — GEO Health Check for cuecards

Audit cuecards' Generative Engine Optimisation (GEO) layer for accuracy and completeness. GEO = making Cue easy for ChatGPT, Perplexity, Google AI Overviews, and similar systems to understand, summarise, and cite correctly.

What to check

1. llms.txt accuracy (root)

• Version/install command is current: grep "npm install" llms.txt matches current npm package
• Supported agent count matches docs/llms-full.txt

2. web/public/llms.txt accuracy

• Consistent with root llms.txt
• All linked GitHub URLs resolve (spot-check 3)

3. content/geo/entity-profile.json

• Valid JSON: python3 -m json.tool content/geo/entity-profile.json > /dev/null
• @type includes SoftwareApplication
• featureList count ≥ 8
• keywords list ≥ 8 terms
• description field matches README tagline (no drift)

4. content/geo/brand-facts.md

• Agent count in "Supported agents" section matches docs/llms-full.txt
• Install commands are syntactically valid bash
• Token cost claim (10–25×) matches README claim

5. content/geo/answer-engine-faq.md

• Every answer is ≥ 2 sentences (too short = not citation-worthy)
• Install commands match current package.json package name
• No questions reference outdated file names (e.g. .cue-profile should now be .cue.profile)

6. content/geo/ai-discovery-map.yml

• All url: values use the correct GitHub org (opencue/cuecards)
• No dead paths referenced (check do_not_scrape list is still accurate)

7. Consistency cross-check

• Product name: "cuecards" (lowercase) used consistently — not "CUE" or "Cue" as primary
• npm package: cue-ai consistent across all geo files
• Repository: https://github.com/opencue/cuecards consistent

Report format

Flag any factual drift (e.g. agent count changed, install command changed) as HIGH priority — stale GEO content is worse than no GEO content.