AI Stack Auditor

The average knowledge worker now pays for three to five AI subscriptions. Most have overlap they don't realise — Claude Pro and ChatGPT Plus both doing general chat, Cursor and Copilot both doing IDE completion, Perplexity and ChatGPT-with-browsing both doing research. Meanwhile they're missing tools that would genuinely help, because they assume they're "already covered." This skill fixes both problems at once.

The process starts with an inventory: name, tier, monthly cost, what they actually use it for (not what it can do), and how often. The "if this disappeared tomorrow, what would break?" question surfaces cut candidates fast. Then it builds a capability matrix — chat, code, research, image generation, document analysis, automation — to expose the overlapping cells (redundancy) and the empty rows (gaps). Underutilised tools get a cost-per-use number: a £20/month tool used twice a month is £10 per use, and that figure is harder to argue with than "you barely use this."

The output is a Keep / Cut / Downgrade / Add / Consolidate table with specific reasons per line. Free tiers are treated as legitimate — if Perplexity's free tier covers your usage, the skill says so rather than upselling you to Pro. New tools come with a trial-before-committing rule rather than a straight paid recommendation. And it factors in switching costs: a tool the team has built workflows around costs more to cancel than the subscription delta, and the audit accounts for that.

What it deliberately doesn't do: recommend tools to justify its own existence. If your stack is already lean and effective, you're told that. If you're on a free tier and your usage doesn't justify paying for anything yet, you're told that too. The principle is the one most reviews break — the best tool is the one you'll actually use, not the theoretically superior one you forget about.