known-errors

# Known Errors & Recovery When a Creatify tool call errors, match it to one of these and recover accordingly. Do not blindly retry, and do not start debugging the repo or local services — these are access/state/quota conditions, not code bugs. ## Auth / access - Symptom: unauthorized, or a tool says it needs authentication. - Recover: run `codex mcp login creatify` and retry. Confirm the account is the one that owns the target brand/product. ## Wrong brand space - Symptom: a product/brand/folder "not found" even though the user says it exists. Some tools say it explicitly — e.g. `asset_get`'s "exists but is in a different workspace" — which confirms this case. - Recover: the data likely lives in another brand space. `brand_list`, then `brand_switch` to the right one, and retry with the correct id. If it spans accounts (not just brand spaces), re-authenticate under the right account. See `product-and-brand`. ## Quota / plan gating - Symptom: voice-slot full or a creation-time cap (voice cloning); a model rejected by `plan_level` (asset generation). - Recover: surface the limit to the user (they may need to free a slot or upgrade) instead of retrying. Offer a lower-tier model when one fits. ## Not-found / bad id - Symptom: an id-scoped call fails. - Recover: use ids returned by list/get tools (`avatar_list`, `voice_list`, `product_list`, `asset_list_models`, session results); never guess ids. ## Long-running / "still processing" - Symptom: a `*_get` poll isn't done yet. - Recover: keep polling the matching status tool; a not-yet-done state is not a failure. Only conclude failure on a terminal failed state. See `verification`. ## Missing source media - Symptom: a workflow needs a photo/video/audio the agent doesn't have. - Recover: check whether it already exists (e.g. in the brand/product or project) before asking. If genuinely missing, get it in via `media-upload` or ask the user to provide it. If a host policy blocks the upload, tell the user to upload directly in the Creatify web app.