pm init writes two status registries that disagree: settings.json's embedded schema.statuses omits in_progress's alias while schema/statuses.json declares it #828

pm --version 2026.7.29. Surfaced by a code reviewer on a fresh tracker and confirmed to be systematic.

What happens

pm init writes the status registry in two places, and they disagree about aliases.

.agents/pm/settings.json carries an embedded schema.statuses array and a schema.files.statuses pointer to schema/statuses.json. For in_progress, the canonical file declares an alias and the embedded copy does not:

settings.json  schema.statuses:  {"draft":null,"open":null,"in_progress":null,       "blocked":null,"closed":null,"canceled":["cancelled"]}
schema/statuses.json:            {"draft":null,"open":null,"in_progress":["in-progress"],"blocked":null,"closed":null,"canceled":["cancelled"]}

Note canceled does carry ["cancelled"] in both — so the embedded copy is not simply alias-free, it is alias-incomplete. That asymmetry is what makes it look like a bug rather than a deliberate projection.

Systematic, not local

Checked four independently initialized trackers of different ages (three fleet package repos plus a companion repo):

All four agree with each other and disagree internally, so this is what pm init emits rather than drift introduced by a user.

Why it matters

Not a functional break as far as I can tell — pm update <id> --status in-progress resolves fine, so the canonical file is evidently the one the resolver reads. The cost is review noise and a bad repair incentive:

• A reviewer comparing the two files reports it as a data-integrity finding (this is exactly how I found it).
• The obvious "fix" is to hand-edit settings.json to add the alias. That is the wrong move — settings.json is history-audited, and the next pm init or pm schema mutation regenerates it anyway — but nothing in the file says so.

Suggested fix

Either make the embedded schema.statuses block a faithful mirror of schema/statuses.json (including aliases), or omit aliases from the embedded copy entirely for all statuses so it reads as a deliberate summary rather than an incomplete duplicate. The current state — one status with aliases, another without, for no visible reason — is the only version that reads as a defect.

If the embedded block is a cache rather than a source, saying so in a comment or a "_generated"-style marker would stop reviewers (human and automated) from filing this repeatedly.