复制安装命令
用 Codex 或 Claude 安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它先审查 Skill 页面再帮你安装。
复制前请先查看来源、License 和安全提示。
SkillAtlasSkill 详情
jetson-derive-carrier
Official, NVIDIA-verified Agent Skills for Claude Code, Codex, and other coding agents.
审核状态：已审核Quality 72Security 70
项目 README
来源文件：README.md
NVIDIA Agent Skills
Official, NVIDIA-verified Agent Skills for Claude Code, Codex, and other coding agents.
📖 Docs: docs.nvidia.com/skills · 📺 Livestream: From Vulnerable to Verified · 📝 Blog: NVIDIA Verified Agent Skills: Capability Governance for AI Agents
Skills are portable instruction sets that teach AI agents how to use NVIDIA software optimally: Physical AI and robotics workflows, simulation, CUDA-X libraries, RAG and AI Blueprints, and platform tools. This repository is a catalog: skills are maintained in their respective product repos, and mirrored here daily via an automated sync pipeline. Skills are being added continuously, so check back for updates. We are building this infrastructure in the open, and contributions are welcome. See the Roadmap for what is planned next.
Quickstart
Install NVIDIA skills with the default skills CLI flow:
npx skills add nvidia/skills
The CLI runs through npx and prompts you to choose a skill and install destination. You do not need to clone this repo or copy skill folders by hand.
Requires a current
skillsCLI (v1.5.16 or newer). Installing vianpx skills@latest add nvidia/skillsalways uses the latest. On older CLIs (v1.5.15 and earlier), skills may install but not appear in Claude Code — see Troubleshooting.
The skill is available the next time your agent loads skills and encounters a relevant task. For example, ask your agent to "solve a linear programming problem with cuOpt" and the skill guides it through the cuOpt Python API. In Claude Code, run /reload-skills to load newly installed skills in your current session.
Install One Skill Without Prompts
Use this when you already know the skill name and want to skip prompts.
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --yes
Replace cuopt-numerical-optimization-api with any skill name from the Skill Catalog.
Install for a Specific Agent
Use --agent to target a specific AI coding agent. Initially, we'll support common client targets, expanding the list over time. For the full list of clients supported by the spec, see the skills CLI Supported Agents table.
Claude Code
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent claude-code
Codex
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent codex
Snowflake CoCo
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent cortex
Cursor
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent cursor
Kiro
npx skills add nvidia/skills --skill cuopt-numerical-optimization-api --agent kiro-cli
Use --agent more than once to install the same skill into multiple agents.
npx skills add nvidia/skills \
--skill cuopt-numerical-optimization-api \
--agent claude-code \
--agent codex \
--agent cursor \
--agent kiro-cli
Keep Skills Up to Date
New skills land continuously, and existing ones are revised, renamed, or consolidated as the catalog evolves. Refresh what you have installed with:
npx skills update
Run it interactively and the CLI also flags skills that were removed or merged upstream (for example, when several skills are consolidated into one) and offers to remove the stale local copies. Use npx skills list to see what is installed and npx skills check to preview what is out of date first.
Browse the Catalog
Use this when you want to see available NVIDIA skills before installing anything.
npx skills add nvidia/skills --list
For non-interactive installs, global installs, agent-specific installs, updates, removals, and fallback manual copying, see Advanced installation.
Skill Catalog
Getting Help & Contributing
Where to file an issue depends on what's broken:
- Skill content issues (a specific skill has a bug, missing functionality, or incorrect content) — file in the source repo for that product, using the per-product table below.
- Catalog issues (catalog README errors, sync workflow problems, distribution channels, signing/verification flow, docs in this repo) — file here using the catalog issue templates: Bug Report, Feature Request, or Documentation Request or Correction.
- Questions or general discussion — use Discussions. The issue tracker is reserved for bug reports, feature proposals with a design, and documentation issues.
- Security vulnerabilities — follow the disclosure process in SECURITY.md; do not open a public issue.
Per-product source repo links:
For issues with this catalog repo itself (README, structure, listing a new product): open an issue here.
Verifying Skills
Every published skill ships with a detached OMS signature (skill.oms.sig). The sync pipeline drops any skill missing the required artifacts before publishing, so every skill in the catalog carries:
SKILL.md— the skill instructions consumed by the agentskill-card.md— skill identity and governance cardskill.oms.sig— detached OMS signature (verifiable againstnv-agent-root-cert.pem)- A Tier-3 evaluation dataset — accepted at
evals/evals.json,evals/*.json,eval/*.json, orbenchmark/evals.json BENCHMARK.md— generated benchmark report capturing verifiable uplift data
Verify a skill against the NVIDIA trust anchor nv-agent-root-cert.pem:
pip install model-signing
model_signing verify certificate SKILL_DIR \
--signature SKILL_DIR/skill.oms.sig \
--certificate_chain nv-agent-root-cert.pem \
--ignore_unsigned_files
A successful verification confirms that the skill contents have not been modified since signing by NVIDIA.
See Verify Signed Agent Skills for signature layout, the trust pipeline, and policy options.
Roadmap
- ✅ Public skills catalog with NVIDIA-verified skills across multiple products
- ✅ Automated sync pipeline with skills mirrored from product repos daily
- ✅ Security scanning for all published skills covering instruction safety and supply-chain integrity
- ✅ Skills signing so every published skill carries a verifiable NVIDIA signature
- ✅ Skills universal evaluation criteria and task-specific criteria
- ✅ Skill Card with machine-readable metadata for identity, provenance, quality, and behavioral boundaries
- ✅ Sync-time compliance gates — signature drift detection and missing-artifact enforcement
- ✅ Syndication to external marketplaces — Skills.sh, Codex plugin, Claude Code plugin, ClawHub, Hermes Hub
- 🔲 Syndication to additional MCP hubs and partner channels
Repository Structure
NVIDIA/skills/
├── skills/ # NVIDIA-verified skills (count grows continuously),
│ │ synced from upstream product repos
│ ├── README.md # Browser-facing install guidance
│ ├── <product-prefix>-*/ # Flat layout — one dir per skill, product-prefixed
│ │ # e.g. aiq-*, cuopt-*, cupynumeric-*,
│ │ # dali-*, deepstream-*, dicom-*, digital-health-*,
│ │ # dynamo-*, earth2studio-*, holoscan-*, hsb-*,
│ │ # jetson-*, launch-nemo-rl, mcore-*,
│ │ # nemo-automodel-*, nemo-data-designer-plugin,
│ │ # nemo-evaluator-plugin, nemo-mbridge-* (20 skills),
│ │ # nemo-retriever, nemo-rl-* (4 skills),
│ │ # nemoclaw-user-guide, nemotron-*, nemotron-speech,
│ │ # nv-* (medical AI), physicsnemo-*, rag-*,
│ │ # skill-card-generator, tao-*, tilegym-*,
│ │ # vss-* (15 skills), accelerated-computing-cudf,
│ │ # cudaq-guide, portfolio-optimization
│ ├── omniverse-*/ # Physical AI — manually staged (see manual-components.yml)
│ └── physical-ai-*/ # Physical AI — manually staged
├── components.d/ # Product registry — one file per component, teams onboard here
│ ├── README.md # Schema and onboarding instructions
│ └── <product>.yml # one file per registered product
├── plugins/ # Packaged plugin distributions
│ └── nvidia-skills/ # Curated NVIDIA skills bundle (Claude Code, Codex)
├── plugins.d/ # Plugin build registry — config for `build-plugins.py`
│ ├── README.md
│ ├── _defaults.yml
│ └── nvidia-skills.yml
├── .claude-plugin/ # Claude Code marketplace metadata
│ └── marketplace.json
├── .agents/plugins/ # Agent marketplace metadata (other clients)
│ └── marketplace.json
├── docs/ # Long-form documentation (published via Fern)
│ ├── README.md # How to build the docs locally
│ ├── index.mdx
│ ├── advanced-install.mdx
│ ├── agent-skill-trust-pipeline.mdx
│ ├── release-checklist.mdx
│ ├── scanning-agent-skills.mdx
│ ├── signing-agent-skills.mdx
│ └── skill-cards.mdx
├── fern/ # Fern docs site configuration
├── .github/
│ ├── workflows/ # Sync pipeline, plugin validation, DCO check, author verify
│ └── scripts/ # regenerate-readme.sh, build-plugins.py,
│ # manual-components.yml (temp Physical AI catalog
│ # exception, removed after Computex 2026),
│ # marketplace/metadata.json (skill metadata sidecar)
├── nv-agent-root-cert.pem # Trust anchor for OMS signature verification
├── skills.sh.json # Skills.sh marketplace grouping config
├── CHANGELOG.md
├── CONTRIBUTING.md # Contribution guidelines
├── SECURITY.md # Security reporting policy
├── CODE_OF_CONDUCT.md # Community code of conduct
├── LICENSE-APACHE # Apache 2.0 (source code)
└── LICENSE-CC-BY-4.0 # CC BY 4.0 (documentation/skills)
Skills are maintained in their respective product repos (see the Source column in the Skill Catalog) and synced to this repo daily. Products only appear under skills/ after the sync pipeline confirms each skill carries:
skill.oms.sig— detached OMS-format signature (verifiable againstnv-agent-root-cert.pem)skill-card.md— skill identity and governance card- A Tier-3 evaluation dataset — accepted at
evals/evals.json,evals/*.json,eval/*.json, orbenchmark/evals.json
When evaluation runs produce a BENCHMARK.md, it ships alongside the skill so consumers can see verifiable benchmark uplift data.
Standards & Compatibility
This repository adheres to the Agent Skills specification:
- Skills are portable directories with a
SKILL.mdfile at their root. - Metadata uses YAML frontmatter with required
nameanddescriptionfields. - Skills follow a progressive disclosure model — lightweight metadata loads at startup, full instructions load on activation.
- Validate your skill using the
skills-refreference library.
License
Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
This code is dual-licensed with documentation/skills under the CC-BY-4.0 AND source code under Apache-2.0 license terms. The full license texts can be found in LICENSE-APACHE and LICENSE-CC-BY-4.0 respectively.
其他
中风险
- 来源需自行核对维护者身份。
- 包含脚本或命令调用，安装前请复核。
- 可能需要外部 token、网络权限或第三方服务。
- 未检测到高风险命令。
- 扫描发现：1 条。
Codex — Git Clone 安装
- 安装前请先查看来源仓库和风险报告。
- 克隆仓库：git clone https://github.com/NVIDIA/skills.git
- 将 "skills/jetson-derive-carrier" 文件夹复制到 Codex 的 skills 目录中。
- 重启 Codex 让新的 skill 生效。
Codex — 手动复制安装
- 安装前请先查看来源仓库和风险报告。
- 从源仓库下载 SKILL.md 及相关文件。
- 在 Codex 的 skills 目录中创建新文件夹。
- 将所有 skill 文件复制到新文件夹中。
- 重启 Codex 让新的 skill 生效。
Claude Code — Git Clone 安装
- 安装前请先查看来源仓库和风险报告。
- 克隆仓库：git clone https://github.com/NVIDIA/skills.git
- 将 "skills/jetson-derive-carrier" 文件夹复制到 Claude Code 的 skills 目录中。
- 重启 Claude Code 让新的 skill 生效。
Claude Code — 手动复制安装
- 安装前请先查看来源仓库和风险报告。
- 从源仓库下载 SKILL.md 及相关文件。
- 在 Claude Code 的 skills 目录中创建新文件夹。
- 将所有 skill 文件复制到新文件夹中。
- 重启 Claude Code 让新的 skill 生效。
Cursor — Git Clone 安装
- 安装前请先查看来源仓库和风险报告。
- 克隆仓库：git clone https://github.com/NVIDIA/skills.git
- 将 "skills/jetson-derive-carrier" 文件夹复制到 Cursor 的 skills 目录中。
- 重启 Cursor 让新的 skill 生效。
Cursor — 手动复制安装
- 安装前请先查看来源仓库和风险报告。
- 从源仓库下载 SKILL.md 及相关文件。
- 在 Cursor 的 skills 目录中创建新文件夹。
- 将所有 skill 文件复制到新文件夹中。
- 重启 Cursor 让新的 skill 生效。
GitHub Copilot — Git Clone 安装
- 安装前请先查看来源仓库和风险报告。
- 克隆仓库：git clone https://github.com/NVIDIA/skills.git
- 将 "skills/jetson-derive-carrier" 文件夹复制到 GitHub Copilot 的 skills 目录中。
- 重启 GitHub Copilot 让新的 skill 生效。
GitHub Copilot — 手动复制安装
- 安装前请先查看来源仓库和风险报告。
- 从源仓库下载 SKILL.md 及相关文件。
- 在 GitHub Copilot 的 skills 目录中创建新文件夹。
- 将所有 skill 文件复制到新文件夹中。
- 重启 GitHub Copilot 让新的 skill 生效。
Windsurf — Git Clone 安装
- 安装前请先查看来源仓库和风险报告。
- 克隆仓库：git clone https://github.com/NVIDIA/skills.git
- 将 "skills/jetson-derive-carrier" 文件夹复制到 Windsurf 的 skills 目录中。
- 重启 Windsurf 让新的 skill 生效。
Windsurf — 手动复制安装
- 安装前请先查看来源仓库和风险报告。
- 从源仓库下载 SKILL.md 及相关文件。
- 在 Windsurf 的 skills 目录中创建新文件夹。
- 将所有 skill 文件复制到新文件夹中。
- 重启 Windsurf 让新的 skill 生效。
查看 SKILL.md 原文
name: jetson-derive-carrier
description: >-
Bootstrap a custom carrier board by forking carrier files and
scaffolding a DT overlay from the reference devkit. Use after
jetson-init-source; not for module-level or kernel-DTB changes.
version: 0.0.1
license: "Apache-2.0"
metadata:
data-classification: public
author: "Jetson Team"
tags:
- target-platform
- custom-carrier
- bring-up
- setup
domain: metajetson-derive-carrier
Customize first-run gate for any customize-* skill on a custom
carrier. Resolve the active target per
target-platform-contract.md.
Refuse if no custom_carrier: block, or if
<source.root_path>/Linux_for_Tegra/.git is missing
(jetson-init-source first). Template file create/copy rows land in a
single overlay-tracker commit and copy directly to the
custom-carrier-renamed target path — reference-named pristine files
are never staged (see Fork plan for the local rule, which
overrides the standard pristine + customization split from the
workflow);
conf-chain discovery follows
per-board conf dispatch.
Kernel base DTB is not forked — carrier deltas layer as a DT
overlay wired via OVERLAY_DTB_FILE.
Identifiers (from active profile): <chip> from
reference_devkit.module.id via
catalogue (unknown →
warn, fallback tegra234); <module-id>/<module-sku> from
reference_devkit.module; <carrier-id>/<carrier-sku> from
reference_devkit.carrier; <custom-id>/<custom-sku>/
<custom-flash-conf> from custom_carrier. <real-conf> =
readlink <bsp_image.root_path>/Linux_for_Tegra/<reference_devkit.flash_config>
(NVIDIA convention <devkit>.conf → <carrier>-<module>-a<rev>.conf;
warn + treat top-level as real + skip symlink-wrapper row if not a
symlink).
<custom-id> is a custom carrier board token, not necessarily an
NVIDIA-style pNNNN ID. Use it verbatim in dash-form filenames and DT
compatible strings. When a target file family uses p-stripped numeric
tokens, derive <custom-id-file-token> as NNNN only if <custom-id>
matches ^p[0-9]{4}$; otherwise use <custom-id> unchanged. Do not
reject custom carrier IDs merely because they do not start with p.
Instructions
The fork plan below is the instruction set: one discovery pass, then a
row-by-row fork of the per-board fileset, gated by the commit-message
preview before each git commit.
Fork plan
Every git commit produced by this skill — in the overlay tracker
(Linux_for_Tegra/) or the hardware/ source repo — runs through
the
commit message preview gate.
Surface the staged file list + proposed message to the operator
and require accept / edit / cancel before each git commit; on
cancel leave the index staged for manual resolution.
Materialize renamed forks even when bytes do not change. For every accepted fork-plan row, create and track the custom-carrier target path; do not skip a missing target because the fork is only a rename/copy or is byte-identical to the reference. This includes BCT forks such as pinmux, GPIO/GPIOINT, padvoltage/PMC, misc, and MB2 misc. Skip only when the Acceptance column allows it, the source is an explicit warn-and-skip miss, or the expected target path is already tracked; in the last case, report it as already derived and do not create an empty commit.
Template file create/copy rows squash into one overlay-tracker
commit, and reference-named pristine files are never staged. All
rows that materialize a new (template-derived) file in the overlay
tracker — board flash-conf, flash-conf symlink, MB1 BCT (pinmux,
GPIO/GPIOINT, padvoltage/PMC, misc), MB2 BCT (misc), BPMP DTB (when
opted in), nvpmodel, nvfancontrol — copy directly from
<bsp_image.root_path>/Linux_for_Tegra/<rel>/<reference-name> to
<source.root_path>/Linux_for_Tegra/<rel>/<custom-name> with any
content edits applied before staging. The reference-named
(<carrier-id>-<carrier-sku>-keyed) filename is never staged or
committed in the overlay tracker — only the custom-carrier-renamed
file is tracked. This overrides the standard pristine + customization
split from
commit batching.
All such rows land in a single overlay-tracker commit that adds:
(i) the custom-carrier-named files at their target paths with content
already applied, (ii) the flash-conf symlink, and (iii) the targeted
flash-conf content rewrites inside the renamed flash-conf for
PINMUX_CONFIG / GPIO_CONFIG / GPIOINT_CONFIG / PMC_CONFIG /
MISC_CONFIG / MB2_BCT (plus BPFDTB_FILE when BPMP DTB is opted
in). Rows that edit existing upstream files instead of creating
templates — the nvpower.sh patch — and the overlay wire-up
(OVERLAY_DTB_FILE+= append to the just-created flash-conf fork,
treated as a temporally distinct phase per
commit batching)
remain separate commits per their own rows. The DT overlay skeleton
commits in bsp_sources/hardware/, not the overlay tracker, so it is
unaffected. The commit message preview gate fires once on the
squashed commit; warn-and-skipped rows do not contribute, and if every
row in the bundle is already tracked the commit is omitted entirely.
Discovery (single batched pass)
Run one discovery pass; do not interleave with staging. Do not
truncate listings — every candidate filename in each scanned
directory must be visible to the matcher. head, tail, | head -N,
| tail -N, and any other row-limiting filter are out of bounds for
this step; use ls -1 / find unclipped, or grep on the full output.
Truncating risks false warn-and-skip calls when the matching file
sits past the cutoff. Capture:
- Flash-conf vars:
DTB_FILE/TBCDTB_FILE/BPFDTB_FILE/PINMUX_CONFIG/PMC_CONFIG/GPIO_CONFIG/GPIOINT_CONFIG.DTB_FILE/TBCDTB_FILEare captured for reference only so the overlay wire-up can derive<dtb-stem>for the overlay filename — they are never rewritten by this skill. - BCT
.dtsatbootloader/generic/BCT/;.dtsisiblings live one level up atbootloader/— NOT alongside. - nvpmodel:
nvpmodel_<module-id>_<module-sku>*.conf. nvfancontrol:nvfancontrol_<module-id>_<module-sku>_<carrier-id>_<carrier-sku>.conf. nvpower.shanchors for (a)–(d):<carrier-id>cvb branch,<module-id>-<module-sku>SKUelif,tegra<chip>nvpmodel cascade,tegra<chip>nvfancontrol cascade.
| File / category | Discovery | Acceptance | Fork rule |
|---|---|---|---|
| Board flash conf | <real-conf> | Always | Filename sub <carrier-id>-<carrier-sku> → <custom-id>-<custom-sku>. Content sub is targeted, not blanket: rewrite RHS only for vars whose file this skill forks — PINMUX_CONFIG, PMC_CONFIG, GPIO_CONFIG / GPIOINT_CONFIG, MISC_CONFIG, MB2_BCT. Other carrier-keyed vars (DTB_FILE, TBCDTB_FILE, SCR_CONFIG, PMIC_CONFIG, DEVICEPROD_CONFIG, PROD_CONFIG, MINRATCHET_CONFIG, UPHY_CONFIG, dynamic OVERLAY_DTB_FILE+=) MUST stay at reference values — their files aren't forked here and a blanket sed would point them at nonexistent files. DTB_FILE / TBCDTB_FILE specifically: base DTB is not forked; the overlay row appends a new OVERLAY_DTB_FILE+= line instead. BPFDTB_FILE: opt-in extra commit. Never touch <chip>, <module-id>, xxxx. |
| Flash-conf symlink | Unconditional | Always (skip if <real-conf> not a symlink) | New symlink <custom-flash-conf> → flash-conf fork |
| MB1 BCT pinmux | PINMUX_CONFIG + #include follow ¶ | Always | Rename rule † |
| MB1 BCT GPIO | GPIO_CONFIG/GPIOINT_CONFIG + #include follow ¶ | Always | Rename rule † |
| MB1 BCT padvoltage | PMC_CONFIG + #include follow ¶ | Always | Rename rule † |
| MB1 BCT misc | MISC_CONFIG + #include follow ¶ | Always | Rename rule † |
| MB2 BCT misc | MB2_BCT + #include follow ¶ | Always | Rename rule † |
| Kernel DTB + source DTS | (reference only — see header) | Never forked | Base DTB stays at reference. No pristine binary copy in the overlay tracker; no source DTS fork in bsp_sources/hardware/. Carrier deltas live entirely in the DT overlay (next row). |
| DT overlay skeleton (NEW) | Unconditional | Always | Create <source.root_path>/hardware/nvidia/<chip>/nv-public/overlay/<chip>-<custom-id>-<custom-sku>+<module-id>-<module-sku>.dts (skeleton ‡); commit in hardware/, push to origin |
| Per-dir Makefile registration | <source.root_path>/bsp_sources/hardware/nvidia/<chip>/nv-public/overlay/Makefile | Always | Append dtbo-y += <chip>-<custom-id>-<custom-sku>+<module-id>-<module-sku>.dtbo after the last literal-named dtbo-y += entry (BEFORE the $(addprefix $(makefile-path)/,$(dtbo-y)) prefix block — inserting after dtbo-y += $(old-dtbo) silently drops the .dtbo). Same position-sensitive idiom and snippet as the composite slot's Makefile patch in ../jetson-build-source/references/composite-registration.md#makefile-patch-idempotent-position-sensitive. Commit in hardware/. Without this row, nvidia-dtbs never produces the .dtbo referenced by the next row's OVERLAY_DTB_FILE+= line and flash.sh aborts mid-flash on the missing file. |
| Overlay wire-up | Unconditional | Always | Append OVERLAY_DTB_FILE+=",<chip>-<custom-id>-<custom-sku>+<module-id>-<module-sku>.dtbo" to flash-conf fork (extra commit) |
| nvpmodel config | rootfs/etc/nvpmodel/nvpmodel_<module-id-num>_<module-sku>.conf | Always | Filename: append _<custom-id-file-token>_<custom-sku> |
| nvfancontrol config | rootfs/etc/nvpower/nvfancontrol/nvfancontrol_<module-id-num>_<module-sku>_<carrier-id-num>_<carrier-sku>.conf | Always | Filename: substitute carrier portion (append _<custom-id-file-token>_<custom-sku> if source is module-keyed only) |
nvpower.sh patch | rootfs/etc/systemd/nvpower.sh | Always | Pristine + single customization commit with 4 insertions: (a) cvb cascade elif [[ "${machine}" =~ "<custom-id>" ]]; then cvb="<custom-id-file-token>" before reference; (b) inside module-SKU branch set machine="<module-id>-<module-sku>-<custom-id>-<custom-sku>"; (c) nvpmodel cascade branch for composite machine key → conf_file= nvpmodel fork; (d) cvb-keyed nvfancontrol cascade branch → conf_file= nvfancontrol fork |
| BPMP DTB | prebuilt at bootloader/generic/<BPFDTB_FILE> (alt naming: no p prefix, xxxx SKU) | Opt-in y/N, default N — module-level, usually shared with reference | Fork binary: pristine + filename sub -<carrier-id-num>- → -<custom-id-file-token>- (content unchanged). Extra commit on flash-conf fork rewriting BPFDTB_FILE to the renamed binary. |
¶ #include follow. Scan each captured .dts for #include "..." directives anywhere in the file (NVIDIA BSPs put them
inside BCT node bodies); carrier-keyed includes join the fork list.
Stop at one level.
† Rename rule. Carrier-keyed source (e.g.
…-p3834-xxxx-p4071-0000.dts) → filename sub
<carrier-id>-<carrier-sku> → <custom-id>-<custom-sku>; rewrite the
flash-conf variable (e.g. PINMUX_CONFIG) to the renamed filename in
the same customization commit. Module portion is preserved
verbatim from pristine — if pristine has p3834-xxxx, fork keeps
p3834-xxxx; if pristine has p3834-0008, fork keeps p3834-0008.
Never pin module-SKU on your own. Same rule for any other xxxx
wildcard in the carrier-SKU position when pristine uses it (e.g.
p4071-xxxx → p1234-xxxx). Module-keyed
source (e.g. …-p3767-dp-a03.dtsi) → append
_<custom-id-file-token>_<custom-sku> (underscore-form) or
-<custom-id>-<custom-sku> (dash-form) plus an extra commit on
flash-conf fork rewriting the variable (e.g. PINMUX_CONFIG) to the
suffixed name. Content never substituted by default; prompt if
content holds a literal carrier-id-sku self-reference.
‡ Overlay skeleton. DT plugin overlay (/plugin/;) with one
fragment@0 at target-path = "/"; inside __overlay__ set
model = "<custom_carrier.name> carrier board" and compatible = "nvidia,<custom-id>-<custom-sku>+<module-id>-<module-sku>", "nvidia,<chip>". nvpower.sh machine narrowing reads compatible.
Edit verification
Re-grep after every sed / patch. sed silently no-ops on miss;
exit code 0 proves nothing. Multi-line replacements: use Python
str.replace(old, new), not multi-line sed (brittle to whitespace
drift, silent no-ops). Refuse to commit on verification miss.
Summary. Print forks per repo (count + commit SHAs), BPMP-DTB decision, warn-and-skipped rows (mandatory source file missing on this BSP — don't fail the run), overlay path written.
Examples
Trigger phrases the operator might use:
derive custom carrier
bootstrap custom carrier
fork carrier board from reference devkit
Minimal custom_carrier: block in the active profile that the skill
expects to find:
custom_carrier:
id: p1234 # any token; need not be NVIDIA pNNNN style
sku: 0000
flash_config: p1234.conf
name: Acme Custom Carrier
Purpose
Customize the per-board fileset for a new carrier board so downstream
customize-* skills, nvpower.sh, and the boot chain land on the
custom carrier instead of the reference devkit. Base DTB stays at the
reference; carrier deltas live in a DT overlay so a re-derive against
a new reference BSP only needs to re-run this skill.
Prerequisites
- Active profile resolved per
target-platform-contract.mdwith acustom_carrier:block (id, sku, flash-conf, friendly name). source.root_path/Linux_for_Tegra/.gitinitialized — run/jetson-init-sourcefirst; this skill refuses if the overlay tracker is missing.bsp_image.root_pathpopulated by/jetson-init-imageso the reference flash-conf symlink can be resolved.reference_devkit:populated (module + carrier) so the rename rule has carrier-id/sku tokens to substitute away from.
Limitations
- Kernel base DTB and its source DTS are not forked — carrier
deltas must layer through the DT overlay wired via
OVERLAY_DTB_FILE. - BPMP DTB fork is opt-in (default OFF). Most carrier swaps share the reference BPMP DTB; only opt in when board power/clock topology diverges.
- Flash-conf content substitution is targeted, not blanket. Vars
whose files this skill does not fork (
DTB_FILE,TBCDTB_FILE,SCR_CONFIG,PMIC_CONFIG,DEVICEPROD_CONFIG,PROD_CONFIG,MINRATCHET_CONFIG,UPHY_CONFIG, dynamicOVERLAY_DTB_FILE+=) stay at reference values — rewriting them would point at files that do not exist. - Custom-carrier IDs need not follow NVIDIA
pNNNNstyle; the p-stripped numeric token is derived only when the ID matches^p[0-9]{4}$. - Does not customize module-level files. Use the relevant
jetson-customize-*skill for pinmux, clocks, fan curves, nvpmodel, PCIe, UPHY, USB, MGBE, or camera deltas.
Troubleshooting
- Refuses with "no
custom_carrier:block" — declare the block in the active profile before re-running; the skill will not infer it. - Refuses with ".git missing in
Linux_for_Tegra/" — overlay tracker not initialized. Run/jetson-init-sourcefirst. - "
<real-conf>not a symlink" — NVIDIA convention is a<devkit>.confsymlink to<carrier>-<module>-a<rev>.conf; if the top-level is the real conf, the skill warns, treats it as real, and skips the symlink-wrapper row. Not an error. sedreported success but the edit is missing —sedsilently no-ops on miss. The skill re-greps after every patch and refuses to commit on verification miss; for multi-line replacements switch to Pythonstr.replace.- "chip: unknown" / fallback to
tegra234— module ID is not in the catalogue; update the catalogue rather than override in the profile. - Commit message preview prompt blocks the run — expected gate per commit-batching; accept, edit, or cancel. On cancel the index is left staged for manual resolution.
- A mandatory source file is missing on this BSP — the row is recorded as warn-and-skipped in the summary; the rest of the run still completes.
评论 (0)
暂无评论，成为第一个评论者吧！