@m-kopa/launchpad-cli 0.25.0 → 0.26.1

package/CHANGELOG.md

@@ -6,6 +6,77 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 This project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html);
 pre-1.0 minor bumps may carry breaking changes per ADR 0005.
 
+## 0.26.1 — 2026-06-11
+
+Two `launchpad status` UX faults found live by the owner (fast-track,
+type:bug).
+
+### Fixed
+
+- **Slug inference now reads the local manifest** (Fault 1). Bare
+  `launchpad status` in a directory named e.g. `AI-Audit-Assistant/`
+  containing a valid `launchpad.yaml` (`metadata.name: ai-audit`) used
+  to fail with "slug not provided + cannot infer from cwd" — inference
+  only knew the `launchpad-app-<slug>` directory-name convention and the
+  manifest's own declared slug was ignored. Resolution order is now:
+  (1) explicit positional / `--slug`, (2) the local manifest's slug
+  (`./launchpad.yaml` or `--file`; v1 `metadata.name` / v2
+  `metadata.slug`), (3) the directory-name convention. When the manifest
+  and the directory name disagree, the manifest wins with a stderr note.
+  Applied consistently across every verb that infers (`status`, `pull`,
+  `recover`, `logs`, `review`, `merge`, `envvars`, `destroy`, legacy
+  `deploy`) — the five drifted copies of the helper are centralised in
+  one module (`recover`'s inference was dirname-first; it is now
+  manifest-first like everything else).
+- **`launchpad status <slug>` no longer demands a local manifest**
+  (Fault 2). With an explicit (or inferred) slug and no local
+  `launchpad.yaml`, status used to exit 2 on ENOENT — after already
+  fetching the lifecycle and even saying the app was live in its own
+  error text. It now degrades to the live-truth-only view: lifecycle +
+  deployment block (live/failed/in-progress, trigger, build outcome,
+  standing exceptions) plus a one-line "drift not checked" note, exit 0.
+  `--json` carries `"drift": null` and the `live_drift_unknown` state as
+  the not-evaluated discriminant. `--strict` without a manifest exits 0
+  (drift can't be evaluated; status says so) unless the live build
+  itself failed, which exits 1. The hard exit-2 error remains for a
+  present-but-unreadable or invalid manifest, and the usage error
+  remains when no slug can be resolved at all.
+
+(Note: 0.26.0 was never published to npm — a failed docs gate, since
+fixed — so this release ships 0.26.0's `recover` verb and these fixes
+together.)
+
+## 0.26.0 — 2026-06-11
+
+### Added
+
+- **`launchpad recover [<slug>]`** (sp-devlp1 AC12). Repairs an app whose
+  registry record is stuck at a terminal provisioning failure although the
+  app is actually live (the ai-audit class: a since-fixed platform bug
+  failed the record after content shipped, leaving `launchpad status`
+  short-circuited on the failed lifecycle with no CLI way out). The bot
+  verifies LIVE Cloudflare state first — Pages project exists + a
+  successful production deployment is serving — and only then reconciles
+  the record to `live`, preserving every non-lifecycle field. A not-live
+  app is refused with exactly what was checked and the next steps
+  (`deploy --resume` / `--abandon` / re-deploy): recover never fabricates
+  a live state. Owner/editor/break-glass gated, audited, fail-closed
+  (Cloudflare unreachable → nothing changes). Recovering a healthy app is
+  a no-op success.
+
+### Fixed
+
+- **`launchpad deploy` usage text no longer advertises the removed
+  `--platform-repo` flag** (apply runs server-side via portal-bot). The
+  banner now reflects the real mode surface, and a misused flag (e.g.
+  bare `deploy --yes`) prints a pointed one-line error instead of a wall
+  of stale usage.
+- **Policy-rejected deploys now say the retry is clean.** A
+  bundle-policy / app-boundary / build-command rejection leaves no
+  residue (nothing committed, no slug claimed, no staged bundle) — the
+  error now states that explicitly: fix the file and re-run
+  `launchpad deploy`.
+
 ## 0.25.0 — 2026-06-11
 
 ### Changed