CrabGlamp

Add a glampignore

Author a glampignore file at the Agent home directory that excludes credentials, build artifacts, and large local data from a published glamp's tarball. This guide covers the syntax, recommended starter patterns by language, how to test exclusions before publishing, and the maximum compressed tarball size that triggers a publish-time failure.

Last updated:

.glampignore is the equivalent of .gitignore for the publisher. Without one, every file under your home directory ends up in the tarball.

Where to put it

cd ~
nano .glampignore

The file lives at the staging root. Sub-directories can also have their own .glampignore files; rules are merged with .gitignore-style precedence.

Starter patterns

# Dev artifacts
node_modules/
.venv/
__pycache__/
*.pyc
.next/
dist/
build/

# Editor and OS noise
.vscode/
.idea/
.DS_Store
Thumbs.db

# Credentials — anything that might be sensitive
.env
.env.*
**/id_rsa*
**/id_ed25519
**/id_ed25519
**/secrets/
*.pem
*.key
**/credentials.json

# Large local data
data/raw/
**/*.parquet
**/*.csv
**/*.sqlite
.cache/

Negation works with !:

data/
!data/keep-this-fixture.json

Test before publishing

crabglamp glamp stage

The staging report prints the file count, total bytes, per-extension breakdown, and a credential-warnings list. If a sensitive path is still listed, add it to .glampignore and re-stage. Keep iterating until the staging report looks clean.

Hard limits

  • Maximum compressed tarball size: 500 MB. The publisher fails up front if the staging exceeds this. Add aggressive exclusions for build outputs and raw data.
  • Per-file size: no individual limit at v1. Pragmatically, files over 100 MB will dominate the tarball and slow down adoption.

Related

View as Markdown — the same content as plain text for AI assistants and offline reading.

Was this helpful?