find-skills

Warn

Audited by Gen Agent Trust Hub on Feb 21, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • EXTERNAL_DOWNLOADS (MEDIUM): The skill facilitates the installation of agent skills from external GitHub repositories using npx skills add <owner/repo@skill>. While the [TRUST-SCOPE-RULE] applies to the mention of vercel-labs/agent-skills, the general capability allows the agent to fetch code from any untrusted repository.
  • REMOTE_CODE_EXECUTION (MEDIUM): Installation via npx involves executing remote code. The skill specifically recommends the -y flag (npx skills add <package> -g -y), which bypasses confirmation prompts, potentially allowing an attacker to trick the agent into installing and running malicious code automatically.
  • COMMAND_EXECUTION (MEDIUM): The skill utilizes shell commands (npx skills find, npx skills add) to perform its primary functions. These commands are executed in the local environment and are driven by user-supplied parameters.
  • PROMPT_INJECTION (LOW): This finding identifies an Indirect Prompt Injection surface (Category 8).
  • Ingestion points: User-provided search queries are interpolated into the [query] placeholder.
  • Boundary markers: None identified; queries are passed directly to the shell.
  • Capability inventory: Subprocess execution via npx.
  • Sanitization: None; the skill does not specify escaping or validation for the user query before shell execution.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 21, 2026, 12:21 PM