lore publish

Package and publish a skill to the Lore registry.

Usage


lore publish [<path>] [flags]

Description

Validates the manifest.yaml, compresses the package directory into a .lore archive, and uploads it to the registry. Requires lore login or LORE_AUTH_TOKEN.

Flags

Flag	Description
`--dry-run`	Validate and package without uploading
`--registry`	Registry URL

Examples


# Publish from the current directory
lore publish
 
# Publish from a specific directory
lore publish ./my-skill
 
# Dry run — validate without uploading
lore publish --dry-run

Package Structure

Your package directory must contain:


my-skill/
├── manifest.yaml          ← Required
├── README.md
├── LICENSE
└── content/
    ├── en-US.md           ← Required (default language)
    └── pt-BR.md           ← Optional translations

See Package Authoring for the full manifest.yaml schema.

Published packages are immutable. Once a version is published, it cannot be overwritten. Use a new version number for updates.

Bundling extra files with spec.include

If manifest.yaml declares spec.include, lore push/lore publish also bundles every directory or file it lists into the archive:


spec:
  include:
    - templates


✓ Published m1cloud/dotnet-service:1.0.0

There is no more automatic assets//examples/ bundling — those directories are only included if listed in spec.include. If the CLI finds one of those directories locally but it isn’t declared, it prints a [WARN] and leaves it out of the archive rather than failing the publish.

spec.include content is capped by size (5 MB total / 2 MB per file by default) and the whole .lore archive is capped too (10 MB by default). lore push/lore publish validate this before uploading and fail fast with a descriptive error if a limit is exceeded — the Registry enforces the same limits on ingestion as the final authority.