L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Productivité · SKILL.md

workspace-and-auth

workspace-and-auth

Authenticate and select Lepton workspaces safely, build workspace API clients, and troubleshoot auth, URL, and token state without leaking credentials.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: workspace-and-auth description: "Authenticate and select Lepton workspaces safely, build workspace API clients, and troubleshoot auth, URL, and token state without leaking credentials." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Workspace and Auth

Use this sub-skill when the task is to establish, inspect, switch, or debug the Lepton workspace context before doing any cloud work.

Route here for

  • lep login and lep workspace login flows, including credential strings in <workspace_id>:<auth_token> form.
  • lep workspace list, status, id, url, logout, and remove usage.
  • Local workspace records and current-workspace selection.
  • Programmatic workspace API context with APIClient(workspace_id, auth_token, url, workspace_origin_url).
  • Environment fallback with LEPTON_WORKSPACE_ID, LEPTON_WORKSPACE_TOKEN, LEPTON_WORKSPACE_URL, and LEPTON_WORKSPACE_ORIGIN_URL.
  • Diagnosing expired/invalid tokens, missing current workspace, URL/environment mismatches, and 401/403/404 workspace access failures.

Route elsewhere for

  • Creating, updating, stopping, or deleting endpoints, jobs, pods, Ray clusters, fine-tunes, storage, secrets, ingress, logs, or other live resources after auth is established.
  • Dynamic endpoint invocation with the high-level endpoint Client; use the SDK/client sub-skill for endpoint call patterns.
  • General lep command discovery and mutation-safety policy beyond workspace/auth commands.

Non-negotiable safety rules

  1. Never print, echo, log, save, or quote a raw workspace token unless the user explicitly requests token retrieval as the task itself. If the task only needs auth status, provide redacted status instead.
  2. Treat <workspace_id>:<auth_token> as a secret-bearing credential string. Prefer environment variables, secure prompts, or user-side paste into lep login; do not put raw tokens in persistent notes.
  3. Redact tokens in all summaries and copied command output. The safe display form is only a short prefix/suffix mask such as ab****yz or a generic <redacted-token>.
  4. Verify auth before planning live resource operations. A workspace command being read-only does not make it non-credentialed; status and some list paths can contact the workspace API.
  5. If a user asks to print a token but the stated goal is simply to verify login, refuse the raw-token path and run safer checks such as lep workspace id, lep workspace url, lep workspace list, or the bundled context checker.

Fast path

  1. Inspect context without exposing secrets:

    python scripts/check_workspace_context.py
    
  2. If CLI checks are explicitly allowed, run read-only workspace checks with redaction:

    python scripts/check_workspace_context.py --run-cli --commands list id url
    python scripts/check_workspace_context.py --run-cli --commands status
    
  3. If not logged in, use one of these user-side login forms:

    lep login
    lep login -c '<workspace_id>:<auth_token>'
    lep workspace login -i <workspace_id> -t '<auth_token>'
    
  4. If a custom workspace API URL is required by the token issuer, include it at login time:

    lep workspace login -i <workspace_id> -t '<auth_token>' --workspace-url <workspace_api_url>
    
  5. After auth succeeds, hand off resource-specific work to the appropriate workload, storage/secret/ingress, or SDK sub-skill.

Read next

Skills enregistrées