# Operator check: historical `stack apply` placeholder issue Use this check on a previous apply's output and the project where it ran. The [0.2.1 changelog fix](../CHANGELOG.md#021-2026-10-16) records the historical issue; the 0.3.0 and Unreleased entries describe current behavior. The 0.2.1 npm publish failed and 0.2.2 superseded it. Check the installed version and its `stack apply --help`; a source changelog alone does not establish what is installed. Keep logs and vault output within your trusted environment. 1. **Look for the old signal.** Historical wiring attempted to add the marker `TODO_STACK_APPLY` for each declared secret, first with an unsupported `--rotate` option and then without it. Phantom rejected values in command arguments, so the calls failed and the old `envelopes` count was always 0. An old zero count is a sign of the failed attempt, not proof of stored credentials or of an empty vault. Do not create placeholder values. 2. **Read the current wiring result, if present.** For a readable vault, the changelog gives this example: ```text secrets in vault 0 · missing 1 · webhooks 0 (not in Phantom: OPENAI_API_KEY — store with `phantom add `) ``` `not in Phantom` identifies a missing name, not a stored placeholder. Obtain each required value from its provider and run `phantom add ` in a trusted terminal, replacing `` with the reported name and entering the real value when prompted. Do not put the value in arguments. A present name does not establish that its value works. Provider setup may separately have stored real credentials. 3. **Treat skipped checks as unknown.** The same recipe with an unreadable inventory can report: ```text secrets in vault 0 · missing 0 · webhooks 0 · skipped 1 (phantom-wire skipped: openai) ``` These zero counts cover checked names only; they do not prove the vault is empty or complete. Unknown providers are also skipped. Resolve inventory access using [apply troubleshooting](apply-troubleshooting.md) before deciding which names are missing. A `phantom-wire failed` warning also requires investigation; it is not a successful inventory check. 4. **Check the scope and outcome.** Wiring runs only after at least one provider succeeds, and then checks the whole recipe, including failed or rolled-back providers. `--noWire` skips the inventory check and webhook stubs, with no wiring summary. Missing names alone do not fail apply; provider failures do. Verify required credentials and services separately. Current wiring does not write placeholders or fill missing values. It can create commented webhook stubs even for failed providers or when the vault is unreadable, and a repeat can overwrite them. Inspect provider state and existing files before rerunning apply; the stubs need signature verification before use. See [placeholder history](apply-placeholder-history.md) and [apply troubleshooting](apply-troubleshooting.md) for recovery details.