vibeguard-observe

# VibeGuard Observe This skill uses VibeGuard's local observability commands through the Codex App plugin bridge. Observability is local-first: use repository scripts and local JSONL logs before discussing external telemetry. ## When to Activate - User asks to show a VibeGuard dashboard or GUI. - User asks whether hooks are installed, noisy, slow, warning, blocking, or stale. - User asks for project/global hook stats, health, doctor output, or metrics. - User asks to inspect the VibeGuard website/product page from the plugin. ## Red Flags - **No custom Codex panel** - the documented plugin manifest does not expose a local HTML panel field. Use the generated local dashboard instead. - **Runtime/eval collapse** - hook health and eval quality are separate surfaces. - **Remote telemetry drift** - do not enable Prometheus, Victoria, OpenTelemetry, or hosted Sites by default. - **Raw data leak** - do not paste local logs, prompts, secrets, or full command payloads unless the user explicitly asks and it is safe. ## Checklist - [ ] Resolve the source checkout with `plugins/vibeguard/scripts/vibeguard-plugin.sh repo-dir`. - [ ] Use `dashboard` for a local HTML overview. - [ ] Use `health` for recent hook health. - [ ] Use `stats` for project/global trigger summaries. - [ ] Use `doctor` or `codex-status` for install/capability diagnosis. - [ ] Keep setup changes explicit; do not install hooks while answering observe-only requests. ## Commands From a VibeGuard checkout: ```bas...