The job
You’re evaluating a SaaS product and need to know exactly what you’re agreeing to before you enter a card. The pricing grid looks simple, but the real terms often sit in footnotes, linked legal pages, and cancellation language that decides whether you can leave cleanly.
Why this is hard without Sephir
Without Sephir, you jump between pricing, terms, and billing policy tabs, then paste fragments into a separate chatbot that can’t see page context across tabs. You end up rebuilding the evidence manually and still miss key lines about renewal windows, notice periods, or overage math hidden in dense legal copy with opposite incentives.
How Sephir does it
- Open the vendor’s pricing page and any linked terms, billing, or refund pages in separate tabs.
- Pin the tabs you want audited so the agent only reads the relevant sources.
- Open Sephir with
Cmd+Shift+Sand ask for auto-renew clauses, cancellation constraints, trial conversion rules, and overage triggers. - Watch the timeline run
extractPageText(tab)on each page, thencompareTabsto merge conflicts and overlaps. - Ask for a structured verdict: renewal trigger, notice deadline, fee exposure, and exact quoted evidence lines.
- Save the trace as
/audit-pricingso the next vendor check runs from one slash command.
The skill behind it
This skill is a repeatable legal-surface audit: read all visible pricing and policy text, cross-reference contradictions, then return a compact risk summary with evidence.
skill: /audit-pricing
tools: extractPageText, openTab, compareTabs
intent: Audit pricing and legal tabs for auto-renew, cancellation, and overage risk.
What it costs
Sephir runs this on your own ChatGPT Plus (via Codex OAuth) or your own API key. Typical usage is ~3,000–8,000 input tokens and ~500–1,500 output tokens per run on Claude Opus 4, GPT-5, or Gemini 3 Pro. Pricing details for Sephir itself are on /pricing: Free is forever, and Pro is one-time lifetime.