Skip to main content
Skill commands cover direct Registry lifecycle and repository-managed skill dependencies. Start with validate or read commands before changing Registry state.

Commands

list

Lists visible skill artifacts in one workspace or across the account. Usage

get

Gets the latest or a selected version of one skill artifact. Usage
Shows the latest version unless --version is provided.

versions

Lists every retained version for a skill ID. Usage

runs

Shows governance actions recorded for a skill. Usage
JSON output can include action_name, attempt, gate_outcome, id, run_status, target_kind, and version_ordinal.

validate

Checks a local skill directory against server rules without contacting the Gateway. Usage
Reports every validation problem in one run.

push

Creates a Registry skill or appends a version from a local directory. Usage
The command creates a skill or appends a version. Unchanged content is not uploaded again.

pull

Downloads the latest or a selected skill version into a local directory. Usage
The destination defaults to a directory chosen from the skill. Pull stops when the target contains unpushed changes unless --force is supplied. Associates a local directory with an existing Registry skill. Usage
link associates a directory with an existing Registry skill so later pushes append versions. It compares local and remote contents first; --confirm accepts a mismatch without prompting. Removes the local association without deleting files or Registry data. Usage

install

Installs the repository’s declared skill dependencies. Usage
Finds .atlan/package.json, clones the token-scoped marketplace, verifies registered digests, copies selected skills to .atlan/skills, writes the lock file, and creates configured coding-agent links. [dir] defaults to the current directory.

publish

Publishes repository-managed skills to their configured workspaces. Usage

hooks

Installs or removes the repository’s Atlan-managed publishing hook. Usage
install adds the configured pre-push publisher and refuses to overwrite unrelated hooks or shared hook directories. uninstall removes only the Atlan-managed hook.

rollback

Restores an earlier skill version by appending its contents as a new version. Usage
Rollback appends a new version with the selected contents. It does not rewrite or delete later versions.

approve

Clears the review flag on a security-blocked skill. Usage
Inspect skill runs <id> first. Approval does not remove the finding.

delete

Archives a skill while retaining all versions. Usage
Structured artifact reads support --json, --jq, and --template. Valid artifact fields are artifact_status, created_at, description, id, is_latest, is_system, kind, metadata, name, source_digest, updated_at, version_ordinal, and workspace_id.

Examples

Output and behavior

  • Read commands return tables by default and exit 0 for empty results.
  • push, rollback, approve, and delete prompt before changing Registry state unless their explicit confirmation option is supplied.
  • A blocked upload remains inspectable through runs and can require approval before use.
  • Repository install and publish use the same OAuth or ATLANAI_TOKEN credential as other CLI commands.
  • --force, --auto-approve, --confirm, and --prune remove safety checks; use them only after resolving the exact target and expected change.