Read the last day of service logs, fix what's failing and stage the fix.
First, confirm Yard is set up on this machine. Run `yard version`, and check that the Yard skill exists at ~/.agents/skills/yard/SKILL.md (%USERPROFILE%\.agents\skills\yard\SKILL.md on Windows). If either the CLI or the skill is missing, fetch https://yard.sh/setup.txt and follow it through to the end before you do anything else. If both are already in place, go straight to the task below. --- Check my Yard-hosted services for errors from the last day, fix what's causing them, and stage the fixes. 1. In the Yard project linked to this directory, run yard service logs --since 24h --limit 500 --json, once per service with --service <name> if there are several. Group the exceptions and failed requests by cause and show me a short summary with counts. 2. For each cause, find the responsible code and reproduce it with yard dev --json running in the background. If it depends on who's signed in, pick a persona with the yard_dev_identity cookie. The local errors show up at /__yard/dev/api/logs. 3. Fix each one with the smallest change that addresses the cause, and confirm the same request now succeeds locally. If it's a data problem, inspect it read-only with yard db query, and never write to the hosted database without asking me. 4. Run yard service check and yard push, and give me the review URL. If nothing in the logs is an error, say so and stop. Ask me before you publish the fix.