jetson-set-target

por nvidia

Cambia el puntero de la plataforma objetivo Jetson activa a un perfil YAML existente. Úsalo antes de personalizar/construir/flashear para cambiar el objetivo; no para crear perfiles —…

npx skills add https://github.com/nvidia/skills --skill jetson-set-target

Set Target Platform

Overview

This skill is the switcher side of the target-platform contract in ../../context/target-platform-contract.md. It only edits target-platform/active_target.yml; it never authors a new profile YAML. To author a new profile, use the sibling skill jetson-init-target.

Single output: target-platform/active_target.yml updated to point at the user's chosen existing profile.

When to invoke

  • The user asks to switch / change / activate a different target platform among profiles that already exist.
  • A downstream skill refused with "no active target" and target-platform/ already contains one or more profile YAMLs.
  • The user names a target by filename or product, and that profile YAML is already on disk.

If target-platform/ is missing, empty, or contains only active_target.yml, stop and tell the user to run jetson-init-target first.

Procedure

Survey the profiles directory

  1. Read target-platform/active_target.yml (treat a missing file as active: NA for the purposes of this skill — the file will be created in the "Update the active pointer" step if needed, preserving the template header).
  2. List target-platform/*.yaml, excluding active_target.yml itself.
SituationAction
No profile YAMLs existRefuse. Tell the user to run jetson-init-target to author one.
Exactly one profile, already activeTell the user it is already active; no change needed. Stop.
Exactly one profile, not activeConfirm with the user, then jump to the "Update the active pointer" step with that profile selected.
Multiple profilesProceed to the "Show the list" step.

Show the list

Print the profiles in alphabetical order, numbered. Mark the current active selection with (active) so the user knows what they're switching from. For each profile, parse its YAML and show a one-line summary so the user can identify it without opening files: the reference_devkit.name (or custom_carrier.name if present) plus the flash_config.

Sample shape:

Profiles in target-platform/:
   1) acme-vision-x1.yaml          — Acme Vision X1 (custom carrier on AGX Thor T5000) — acme-vision-x1.conf
   2) jetson-agx-orin-32gb.yaml    — Jetson AGX Orin 32GB                              — jetson-agx-orin-devkit.conf
   3) jetson-agx-thor-devkit.yaml  — Jetson AGX Thor T5000                             — jetson-agx-thor-devkit.conf  (active)
   4) jetson-orin-nano-8gb.yaml    — Jetson Orin Nano 8GB                              — jetson-orin-nano-devkit.conf

If a profile YAML fails to parse or is missing the expected fields, show the bare filename with a (unparseable — fix or remove) marker and skip to the next; do not refuse the whole skill over one bad file.

Prompt for selection

Prompt:

Pick a number, the bare filename, the product name (substring match), or cancel.

Resolve to a single profile filename. If the user picks the currently active profile, tell them so and stop without rewriting the pointer. If ambiguous, list candidates and ask again.

Update the active pointer

Edit target-platform/active_target.yml:

  • Set active: <chosen-filename>.yaml.
  • Preserve the header comment block verbatim — only modify the active: line.
  • If the file does not exist (it can be missing if the user hand-set up target-platform/ without the bootstrap step), create it from ../../references/active_target_template.yaml and then set active:.

Confirm

Print a summary: the previous active value, the new active value, the chosen profile's reference_devkit.name (and custom_carrier.name if present), and a reminder that downstream skills (jetson-customize-clocks, jetson-customize-nvpmodel, jetson-customize-fan, jetson-optimize-memory) will now resolve to this target.

If a downstream skill triggered this run, tell the user to re-issue their original request; do not silently re-trigger it.

Prerequisites

  • target-platform/ exists and contains at least one *.yaml profile (excluding active_target.yml).
  • target-platform/active_target.yml is present, or the template at ../../references/active_target_template.yaml is available so this skill can recreate the pointer.

Limitations

  • Edits only the active: line in active_target.yml; never writes profile YAMLs.
  • Refuses when no profile YAMLs exist — routes the user to jetson-init-target instead.
  • Does not resolve fuzzy product names beyond simple substring match; ambiguous selections are re-prompted.

Troubleshooting

  • "no profiles found" refusal — the directory has no *.yaml other than the pointer; run /jetson-init-target first.
  • Pointer rewritten to a missing profile — should not happen because the chosen filename is verified before writing; if seen, hand-edit active_target.yml back to a valid filename or rerun with the right selection.
  • Active pointer reads NA — legitimate "no selection" state; pick any listed profile and the pointer will be set.

Gotchas

  • No authoring here. If the user names a target whose profile YAML is not on disk, refuse and route them to jetson-init-target. Do not silently fall through to authoring.
  • Don't edit active_target.yml to point at a missing profile. Verify the chosen filename exists in target-platform/ before writing.
  • Preserve active_target.yml's comment block when editing — only the active: line should change.
  • Exclude active_target.yml from the profile list. It lives in the same directory but is the pointer, not a profile.
  • active: NA is the legitimate "no selection" state. Treat it the same as a fresh pointer — don't refuse, just show the list and let the user pick.
  • One-profile shortcut. If only one profile exists, do not skip user confirmation — still confirm before flipping the pointer.

References

Más skills de nvidia

compileiq-debug
nvidia
Úsalo cuando algo esté mal: Search() se cuelga, todas las evaluaciones devuelven INVALID_SCORE, las puntuaciones no mejoran, cada configuración devuelve el mismo número, errores de ptxas…
create-github-pr
nvidia
Crear solicitudes de extracción de GitHub usando la CLI gh. Usar cuando el usuario quiera crear un nuevo PR, enviar código para revisión o abrir una solicitud de extracción. Palabras clave de activación -…
nemoclaw-maintainer-cross-issue-sweep
nvidia
Escanea otros issues abiertos para encontrar aquellos que un PR dado también podría corregir o romper accidentalmente. Genera oportunidades de corrección adyacente y riesgos de contradicción con archivo:línea…
fhir-basics
nvidia
Enseña a los agentes cómo funcionan las APIs de FHIR R4, qué recursos están disponibles, cómo consultarlos con parámetros de búsqueda y cómo analizar correctamente todos los formatos de respuesta…
compileiq-validate-result
nvidia
Usar DESPUÉS de que una Búsqueda haya finalizado y ANTES de reclamar cualquier aceleración o enviar un ACF. Carga el CSV de dump_results, extrae los mejores K candidatos (de un solo objetivo)…
changelog-audit
nvidia
Auditar el CHANGELOG.md de Warp antes de un lanzamiento: recuperar entradas perdidas, ordenar por impacto en el usuario, refinar el lenguaje de las entradas, ajustar saltos de línea y (en modo rama de lanzamiento) incrementar comparación…
maintain-dynamic-plugins
nvidia
Mantener los cargadores de plugins dinámicos de NeMo Relay, manifiestos, SDKs nativos de Rust, protocolo de trabajador gRPC, SDK de trabajador Python, documentación, pruebas y cobertura del flujo de trabajo de lanzamiento
dgx-diagnose
nvidia
Diagnostica problemas comunes de la DGX Station GB300: fallos de CUDA, direccionamiento incorrecto de GPU, errores de contenedores vLLM/SGLang, problemas de estado MIG, errores de NVLink/Fabric Manager,…