---
title: openplanr 1.24.1
description: Published 2026-08-04. Report an invalid .planr/config.json as an actionable error instead of a raw stack trace.
url: https://openplanr.dev/docs/changelog/1.24.1
updated: 2026-08-04
related:
  - https://openplanr.dev/docs/changelog/1.25.1.md
  - https://openplanr.dev/docs/changelog/1.25.0.md
  - https://openplanr.dev/docs/changelog/1.24.0.md
---

# openplanr 1.24.1

Published 2026-08-04.

[Release on GitHub](https://github.com/openplanr/OpenPlanr/releases/tag/v1.24.1)

## Patch Changes

Report an invalid `.planr/config.json` as an actionable error instead of a raw stack trace.

`targets` and `createdAt` are the two config fields required with no default, while every
other field defaults. A config missing either — hand-edited, partially written, or
hand-authored — crashed **every** config-reading command with an unhandled `ZodError`
stack trace, because the CLI's top-level handler rethrows anything without an `E_` code.

The failure now names the file, every failing field with its path, and the command to
repair it, exiting cleanly through the handler's existing error contract. A genuinely
missing required field still fails — it just says so legibly rather than dumping a trace.

Make `setup` and `doctor` describe the install they actually produced.

**`doctor` names the skill file that exists.** Its three `operate-skill` messages hardcoded
the namespaced `planr-operate`, so a `--no-prefix` install — where the file on disk is
`operate` — got a diagnostic with the right status about a filename the user does not have.
All three now interpolate the installed name.

**The setup preview states the naming scheme.** The choice is persisted per project, so a
plain re-run could install bare verbs with nothing in the summary saying so. The preview
now reports `Command names: namespaced` or `bare (--no-prefix)`, resolved from the same
value the installer uses.

**The Cursor no-op is now pinned by a test.** `applyCommandPrefix` is threaded through the
Cursor branch but computes identity there, since no Cursor rule filename starts with
`planr-`. That "intentional symmetry, not dead code" claim rested entirely on a source
comment; a later broadening of the prefix match would have silently begun renaming Cursor
rules with nothing to catch it.

Stop `openplanr upgrade` from offering a downgrade, and make it actually print the plugin-half commands it promises.

**An installed version ahead of the registry is no longer an "upgrade".** Drift was computed
as plain inequality, so "different" and "older" were the same thing: anyone on a build ahead
of published — a linked dev build, a prerelease, a maintainer mid-release — was offered a
downgrade labelled as an upgrade, and accepting "always keep me current" would have rolled
the newer build back on every invocation. Only a version strictly _behind_ now counts.
Range violations are untouched, since those are direction-independent.

**The prescription is no longer promised and withheld.** When the CLI is already current but
the host plugins trail — the state every release creates for anyone who upgrades the npm half
first — `apply` returned early with a message ending "run the prescribed commands below" and
then printed nothing, on both the human and `--json` surfaces. The commands are now built on
that path too, from the same helpers the post-upgrade path uses, so the two can never print
different instructions for the same machine.

**The skills plugin is no longer silently omitted.** The version the CLI targets for the
skills bundle had drifted three releases behind, and because the prescription derives its
target from it, a genuinely stale skills plugin was left out of the commands entirely — a
user could run every prescribed command and still be behind, believing they were current.