Jetson Document Binder
Registers staged Jetson reference documents in the active target profile.
The skill only registers pointers to pre-downloaded documents, does not download files, and requires confirmation for auto-matches; ambiguous matches are not silently bound. However, it mutates the active profile in place without an explicit backup, transactional write, or rollback mechanism, and permits nonexistent paths and unvalidated URLs, so points are deducted.
The workflow, abnormal-input branches, merge behavior, and failure guidance are fairly detailed. However, the referenced contract, templates, catalogue, and YAML dependency are not reproduced in the supplied evidence, and no skill-specific test suite covers the key paths. The static cap is 10, so the score is conservative.
Invocation scenarios, inputs, document slots, and the non-applicable custom-carrier fields are clearly described, and the one-level scan limitation is disclosed. There is no evidence for Chinese-language interaction, precise non-trigger boundaries, or broader environment compatibility, so points are deducted.
The documentation is readable and includes versioning, license metadata, prerequisites, limitations, troubleshooting, references, and rerun guidance. It lacks separate Instructions and Examples sections, the author format is incomplete, the materials present Apache-2.0 and CC-BY-4.0 inconsistently, and maintenance ownership and change history remain insufficiently explicit.
The skill can write user-confirmed document paths into the profile for downstream knowledge-base workflows. However, the benchmark has only two positive tasks, reports 32% Codex effectiveness, and provides no independently reproducible verification of key outputs. Under static calibration the score cannot exceed 7, so it is reduced conservatively.
The supplied benchmark report, task definitions, and expected behaviors provide limited auditable evidence. There is no committed executable test suite, real execution log, or cross-source corroboration for this skill, and the reported conclusions cannot be independently reproduced from the supplied files alone, warranting a low score.
- Add an explicit backup or recoverability mechanism before mutating the active profile, including handling for partial write failure.
- Align the license and author metadata, and add dedicated Instructions, Examples, changelog, and maintenance-owner information.
- Clarify the path policy: the procedure permits absolute paths and URLs, while troubleshooting says documents.* must be relative; these are inconsistent safety boundaries.
- One-level, filename-only globbing can miss documents or produce wrong-product matches; users should verify product, version, and schematic correspondence.
What does this skill do, and when should you use it?
jetson-link-docs updates the documents block of an active Jetson or IGX target-platform profile with paths to reference material already available on disk. It scans a configured document root and suggests matches for developer guides, design guides, schematics, pinmux spreadsheets, and related slots using case-insensitive filename globs. Users must confirm single matches, while multiple matches are presented for explicit selection. The skill only registers pointers; it does not download, copy, or rename documents.
Resolves the active target profile and document-slot schema; resolves documents.root_path, defaulting to workspace/Documents; reads the BSP catalogue Product Token; scans the document root one level deep with case-insensitive globs; prompts users to confirm or select candidates; accepts local paths and URLs; updates the profile's documents block with a round-tripping YAML writer; and prints a summary of auto-bound, manually entered, and skipped fields.
- A Jetson target has been initialized and its downloaded developer guides, schematics, or pinmux files need to be registered.
- An existing target profile needs document references added, changed, or removed.
- jetson-generate-kb reports that no documents are recorded and the profile needs repair.
- A custom carrier requires its schematic or pinmux spreadsheet to be added to the active profile.
- Several document versions are present and the user needs explicit selection to avoid binding the wrong product files.
What are this skill's strengths and limitations?
- Has a narrow, well-defined purpose for the Jetson/IGX profile documents block.
- Combines SKU and product-token filename matching with mandatory user confirmation.
- Supports reruns that merge new bindings while preserving existing ones unless explicitly changed.
- Accepts local paths and URLs and allows optional fields to be skipped.
- Does not download, copy, or rename source documents.
- Automatic scanning is limited to one directory level, so nested archives may require manual selection.
- Requires an active profile, the fixed document schema, and ruamel.yaml or an equivalent round-tripping YAML writer.
- The supplied source provides no test suite, platform support matrix, or performance measurements.
How do you install this skill?
Install the skill with the README-supported skills CLI command: npx skills add nvidia/skills --skill jetson-link-docs --yes. The README does not specify a fixed installation directory; the CLI handles the destination.
How do you use this skill?
Ensure an active Jetson/IGX target profile exists, place reference files under the recorded documents.root_path or the default workspace/Documents, or prepare paths for manual entry. Then ask the agent: "Bind the downloaded Jetson reference documents to the current target profile, scan the document directory, and ask me to confirm each match." The skill prompts for unmatched fields and writes confirmed paths back to the active profile. It does not download documents; without an active profile, use jetson-set-target or jetson-init-target first.