skill-inspector

An agent skill by NVIDIA, from NVIDIA/SkillSpector. Tags: debugging, developer-tools, evaluation, local-workflow.

What it does

Review AI agent skills before installation using NVIDIA SkillSpector and source-aware semantic review. Use when asked whether a skill or downloaded skill folder is safe, trustworthy, installable, over-permissioned, or malicious.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add NVIDIA/SkillSpector --skill skill-inspector

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/NVIDIA/SkillSpector
cp -r SkillSpector/skills/skill-inspector ~/.claude/skills/skill-inspector

Safety box score

Grade B, low risk: 11 of 100 risk points, rated 2026-09-22. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This skill is a review-only markdown guide: it does not ship executable code, and its main actions are reading target skill files, optionally downloading a target repo to a temporary directory, and running the local `skillspector` CLI. The main risks are limited network use for fetching a target skill and the fact that it delegates part of the review to an external CLI, but there is no sign of secrets access, privilege escalation, or hidden execution in the skill itself.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	1 (low)	It directs the agent to run local commands such as `skillspector scan "$TARGET"` and read-only utilities, but it ships no executable scripts of its own.
Network access	1 (low)	It may `clone or download` a user-provided URL into a temporary directory, which is a bounded network action for review purposes.
Filesystem reach	1 (low)	Its file access is scoped to the target skill and a temporary download location, with no instructions to edit or delete outside that scope.
Secrets & credentials	0 (none)	The skill does not request API keys, tokens, browser sessions, `.env` files, or other credentials.
Privilege & persistence	0 (none)	There is no `sudo`, service installation, global configuration, or shell/profile modification behavior.
Instruction hijack surface	0 (none)	The text does not ask the agent to ignore instructions, hide behavior, or follow remote prompt content; it is defensive rather than manipulative.
Irreversible actions	0 (none)	It only produces a review verdict (`APPROVE`, `CAUTION`, or `REJECT`) and does not send messages, deploy, pay, or delete anything.
Transparency	0 (none)	The shipped content is plain Markdown with readable commands and no binaries, obfuscated payloads, or decoded-at-runtime code.

Before installing

Check that `skillspector` on your PATH is the trusted binary you expect before using this skill.
Read `SKILL.md` and confirm there are no bundled scripts or hidden executables; this skill is documentation-only.
If you will review remote skills, test the download step in a scratch temp directory first and inspect the fetched tree before scanning it.

Rated 2026-09-22 by an automated review of 1 file from the skill's repository.

Source

Repository: NVIDIA/SkillSpector (all skills from this repository)
Path: skills/skill-inspector/SKILL.md
Branch: main
Updated: 2026-09-19

sdf — SDFormat/SDF model and world authoring, validation, and simulator handoff.
srdf — MoveIt2 SRDF authoring, validation, and planning-semantics workflow.
urdf — URDF robot description authoring and validation.
agents-get-started — Use when a developer wants to create a new agent project or get started with AgentCore.
c-review — Performs comprehensive C/C++ security review for memory corruption, integer overflows, race conditions, and platform-specific vulnerabilities.
code-improver — Runs an autonomous review-and-fix improvement loop over any code target — a skill, plugin, module, or directory — using a reviewer the user names.