sanitize index-name does not strip all characters Elasticsearch rejects

[MattDevy]

Summary

elastic sanitize index-name lowercases and strips whitespace, but leaves characters that Elasticsearch rejects in index names (e.g. !). The current --help description ("Sanitize an Elasticsearch index or alias name") implies the output is always a valid index name.

Reproducer

$ elastic version
{"version": "0.1.1"}

$ elastic sanitize index-name 'Foo Bar!'
# changes: lowercased, stripped whitespace
foobar!

$ elastic sanitize index-name 'My/Bad:Name*'
# changes: lowercased, stripped forbidden characters
mybadname

The first case still contains !, which Elasticsearch rejects in index names. The second case correctly strips forbidden characters.

Expected behavior

Either:

• Strip every character that Elasticsearch's index-name validator rejects, so the output is always a valid identifier, or
• Document the limited guarantee in --help and the stderr "changes" line (e.g. "removes characters known to be invalid; full validation still required").

Environment

• elastic v0.1.1 (Homebrew on macOS, Darwin 25.5.0)

[JoshMock]

As it turns out, ! is a valid index name character:

❯ node dist/cli.js es indices create --index 'foo!bar2'
{
  "acknowledged": true,
  "shards_acknowledged": true,
  "index": "foo!bar2"
}

❯ node dist/cli.js es indices get --index '*bar2*'
{
  "foo!bar2": {
    "aliases": {},
    "mappings": {},
    "settings": {
      "index": {
        "routing": {
          "allocation": {
            "include": {
              "_tier_preference": "data_content"
            }
          }
        },
        "number_of_shards": "1",
        "provided_name": "foo!bar2",
        "creation_date": "1779213515998",
        "number_of_replicas": "1",
        "uuid": "GJIjYBpFSmaxAGTOxTw7Pg",
        "version": {
          "created": "9060001"
        }
      }
    }
  }
}

The Zod validation was failing but seems to have been fixed by #349. Hard to unravel what was causing what. 😆

Anyway, I'm adding some more exhaustive tests for index-name sanitization in #358 that I pulled straight from the ES repo.

[MattDevy]

As it turns out, ! is a valid index name character:

❯ node dist/cli.js es indices create --index 'foo!bar2'
{
  "acknowledged": true,
  "shards_acknowledged": true,
  "index": "foo!bar2"
}

❯ node dist/cli.js es indices get --index '*bar2*'
{
  "foo!bar2": {
    "aliases": {},
    "mappings": {},
    "settings": {
      "index": {
        "routing": {
          "allocation": {
            "include": {
              "_tier_preference": "data_content"
            }
          }
        },
        "number_of_shards": "1",
        "provided_name": "foo!bar2",
        "creation_date": "1779213515998",
        "number_of_replicas": "1",
        "uuid": "GJIjYBpFSmaxAGTOxTw7Pg",
        "version": {
          "created": "9060001"
        }
      }
    }
  }
}

The Zod validation was failing but seems to have been fixed by #349. Hard to unravel what was causing what. 😆

Anyway, I'm adding some more exhaustive tests for index-name sanitization in #358 that I pulled straight from the ES repo.

TIL! Thanks Josh