replay-vision-scanners-corelisted

# Replay vision scanner mechanics Shared rules for every scanner task. Your task prompt says *which* scanner you create and what you fill in; this skill is *how*. ## Load the authoritative mechanics first Load the scanner skill: `skill-get {"skill_name": "creating-replay-vision-scanners"}`. It owns the create/update mechanics — scanner-type and config shapes, the `RecordingsQuery`, the estimate and quota calls — and the **size-before-you-ship gut-check**: estimate the scanner's monthly **credit** spend, read the org's remaining budget, and compare credit-to-credit. The quota is an org-wide monthly credit budget — never infer it from scanner count, and never compare observation counts against credits. The briefs are deliberately small (bounded sampling, and scoped queries where they have one), so projected spend is normally a tiny fraction of the budget — just create. Only when the credit-to-credit comparison says the spend is a large fraction of (or exceeds) what's left, or the org is already exhausted, ask the user (decline option first): create anyway vs skip. ## Endpoint availability - **If `info vision-scanners-create` says the tool is unknown**: run one `search vision` to confirm, then record a follow-up ("create Replay vision scanners in PostHog once available") and finish the task. - **If every scanner endpoint 404s**: Replay vision is not available for this project — report that in your handoff and finish. Do not retry. - **If a call 403s**: the token lack