skills-cli-usage

Pass

Audited by Gen Agent Trust Hub on Feb 18, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • EXTERNAL_DOWNLOADS (LOW): The skill documentation references the npx skills command, which downloads and executes the skills package from the npm registry. The underlying source code is hosted at github.com/vercel-labs/skills, which is a trusted repository according to the [TRUST-SCOPE-RULE].
  • COMMAND_EXECUTION (LOW): The skill provides various commands for the agent to execute, such as listing, adding, removing, and updating skills. These are standard management operations for the Vercel Skills CLI tool.
  • REMOTE_CODE_EXECUTION (LOW): The npx skills add <source> command allows for the installation of additional skills from arbitrary GitHub repositories or URLs. While this is the intended function of the tool, it constitutes a potential vector for installing untrusted code.
  • PROMPT_INJECTION (LOW): Indirect Prompt Injection (Category 8) surface identified:
  • Ingestion points: The <source> argument in npx skills add commands within SKILL.md allows the agent to ingest content from external, potentially untrusted repositories.
  • Boundary markers (absent): There are no delimiters or explicit instructions to ignore embedded prompts when adding new skills.
  • Capability inventory: The CLI tool has the ability to write files and modify the agent's available toolset/skills.
  • Sanitization (absent): The documentation does not describe any sanitization or validation of the content being added from the external source.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 18, 2026, 09:15 AM