Dentolize · Sandbox Command & Registry Walkthrough
On this pageWho actually uses thisQuestions you might get, and the answersWhat Support cannot fix

For Support

Who actually uses this

This feature's "users" are Dentolize engineers and reviewers, not clinic customers. If a question about /sandbox or "Open sandboxes" ever reaches Support, it's almost certainly from someone internal (a teammate, a contractor doing a review) — not a clinic end user, who has no way to trigger or see any of this.

Questions you might get, and the answers

"I commented /sandbox on my PR and nothing happened." Check comment.author_association — only OWNER, MEMBER, or COLLABORATOR can trigger it (sandbox-command.yml:56-61). Anyone without write access gets a silent 👎 reaction and no reply — that's by design, not a bug. If the commenter does have access and still sees nothing, check whether the comment actually started with /sandbox (leading whitespace or a different first word won't match startsWith, sandbox-command.yml:35).

"It reacted with 😕 instead of doing anything." The sub-command after /sandbox wasn't recognized. The accepted words are: (blank), create, deploy, redeploy, up → redeploy; destroy, down → destroy; reseed; reset-data, reset; docs, update-docs (sandbox-command.yml:66-72). The bot's own reply lists these — point people at that reply first.

"It reacted with 🚀 but no sandbox showed up." 🚀 only means the workflow_dispatch call to Sandbox Ops succeeded (sandbox-command.yml:112-115) — it does not mean the deploy itself finished or succeeded. Check the Sandbox Ops run in the Actions tab for that branch. If it failed there, the sandbox-info PR comment won't appear/update.

"The 'Open sandboxes' issue looks stale / wrong." It refreshes after Sandbox, Sandbox Ops, or the main demo workflow complete, plus every 30 minutes (sandbox-registry.yml:11-17). A brief lag (up to ~30 min) after a park/reap event is expected — that's what the cron schedule is for. If it's been stale much longer than that, check whether the registry workflow itself is failing (Actions tab → "Sandbox Registry").

"I don't see the 🗂 Open sandboxes issue pinned anymore." GitHub allows at most 3 pinned issues per repo. If something else got pinned, this issue could get bumped — the workflow only best-effort attempts to pin and logs (not fails) on error (sandbox-registry.yml:108-116). The issue content itself still updates even if unpinned; check open issues for the 🗂 Open sandboxes title.

What Support cannot fix

The sandbox registry CLI and the persisted GitHub token (/opt/homelab/sandbox/qa-agent/gh-token) both live on the self-hosted runner host, outside this repo. Anything that looks like a problem with the CLI's data (wrong status, missing sandbox) or with authentication (dispatch calls failing with 401/403) needs someone with access to the homelab runner, not a repo-level fix.