Jetson nvpmodel Power Mode Customizer
Safely customize Jetson BSP power modes, frequency caps, and the boot-default profile.
The skill explicitly scopes writes to the overlay, treats the BSP image as read-only, and requires validation of parameters, frequencies, default IDs, propagation targets, and deployment paths. It lacks explicit user confirmation, a complete rollback procedure, and recovery steps for failed sudo copy or flashing operations, so points are deducted. No credential theft, covert exfiltration, or malicious dependency is evident.
Prerequisites, operation flows, format rules, troubleshooting cases, and failure conditions are described in useful detail, making the happy path plausible. However, the workflow depends on unavailable context documents, hardware-specific files, and an external Power Estimator; no committed tests covering key paths or reproducible execution evidence is provided, so the static score remains conservative.
The audience, Jetson/Tegra scope, five operations, boundaries, and routing conditions are clearly stated. Trigger semantics are brief, Chinese-language interaction is not addressed, and non-stock envelope work depends on NVIDIA Power Estimator, which may have mainland-China reachability concerns; points are deducted accordingly.
The documentation has useful layering, examples, limitations, troubleshooting, version, and license fields. However, the skill-card license conflicts with the supplied NOASSERTION metadata, the benchmark reports an author-format finding, and there is no clear changelog, named maintenance responsibility, or update path, so points are deducted.
The skill covers adding, removing, editing, changing defaults, and deployment, with useful propagation and super-configuration safeguards. Its result still depends on the local BSP, DTB, frequency tables, external estimator, and deployment execution; static source material does not independently verify usable outputs, so it does not approach full marks.
The supplied SKILL.md, evaluation files, and report are auditable primary repository material, and the report provides limited evaluation results. There is no committed key-path test suite, independent reproduction record, or multi-source corroboration; the report itself is repository evidence rather than independent verification, so the score is conservative.
- Writes to the overlay can still lead to device-state changes through image promotion and flashing; confirm the target, take backups, and prepare a recoverable image or rollback path before execution.
- Non-stock power envelopes depend on the external NVIDIA Power Estimator; confirm network reachability and independently verify the downloaded file, JetPack release, SKU, and available_frequencies.
- The skill-card license statement conflicts with the supplied NOASSERTION metadata, and the author format has a static validation finding; governance metadata should be reconciled before publication.
- This review did not execute commands or validate hardware behavior, and it does not establish that DTB mapping, propagation updates, or the flashing chain succeed in practice.
What does this skill do, and when should you use it?
This skill targets Jetson/Tegra Orin and Thor platforms and edits per-board nvpmodel configuration in the BSP workflow. It supports adding, removing, editing, and selecting the boot-default power mode. Changes are written to the source overlay rather than the read-only BSP image copy, then committed and deployed. Non-stock power envelopes require NVIDIA Power Estimator, and clock values must match the kernel's available_frequencies table.
Resolves the active target, BSP image, source paths, and SKU-specific nvpmodel file using the kernel DTB compatible string and nvpower.sh mapping; imports pristine files into the overlay tracker; adds or removes POWER_MODEL blocks; edits CPU, GPU, EMC, TPC, and related parameters; changes PM_CONFIG DEFAULT; propagates structural edits to reference, carrier-derived, and _super sibling files; checks parameter declarations, unique IDs, default references, and frequency constraints; and deploys through jetson-promote-image and jetson-flash-image, with a documented direct-copy alternative for a running target.
- A Jetson BSP engineer needs to add a custom power mode for a specific module SKU and make it the boot default.
- A carrier-board developer needs to tune CPU, GPU, EMC, or TPC limits on an Orin or Thor BSP.
- A platform maintainer needs to remove an obsolete mode while updating the boot default and checking hard-coded rootfs references.
- A team needs to propagate one nvpmodel structural change across reference, carrier-derived, and _super configuration files.
- An operator needs a read-only inventory of the modes defined for the active SKU and its boot default.
What are this skill's strengths and limitations?
- Covers the main nvpmodel configuration operations: add, remove, edit, default selection, and listing.
- Clearly separates the read-only BSP image from the writable source overlay and its change history.
- Accounts for DTB mapping, SKU variants, carrier-derived files, and elevated _super envelopes.
- Provides concrete validation rules for declarations, frequency tables, online CPU cores, IDs, and defaults.
- BSP-side only; it does not directly invoke nvpmodel -m on a running device.
- Non-stock envelope changes require NVIDIA Power Estimator; interpolation and extrapolation from existing modes are disallowed.
- Adoption depends on an active target contract, prepared BSP and source repositories, and a matching configuration file.
- The provided SKILL.md does not describe a skill-specific test suite or a validated platform matrix.
How do you install this skill?
Use the installation flow documented in the repository README:
npx skills add nvidia/skills --skill jetson-customize-nvpmodel --yes
To target Codex explicitly:
npx skills add nvidia/skills --skill jetson-customize-nvpmodel --agent codex
The skill becomes available when the agent next loads skills and encounters a relevant task.
How do you use this skill?
After installation, give the agent a request such as:
/jetson-customize-nvpmodel
> list defined power modes
Or:
/jetson-customize-nvpmodel
> set PM_CONFIG DEFAULT to the MAXN profile's ID
Write operations require an active target, a BSP image, and a valid source Linux_for_Tegra Git repository. Changes should be made in the source overlay and committed before running jetson-promote-image and jetson-flash-image. The user must provide the target context and deployment environment; the skill does not blindly infer an unverified configuration filename.