21st-cli

Pass

Audited by Gen Agent Trust Hub on Jul 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for using the 21st CLI tool to perform various project management tasks, such as searching registries, managing teams, and initializing configurations.
  • [EXTERNAL_DOWNLOADS]: The 21st add command downloads React component code from the 21st.dev registry and installs associated npm dependencies using the project's package manager.
  • [DATA_EXFILTRATION]: The 21st publish command uploads local file content and metadata (descriptions, tags, demos) to the 21st.dev platform. This is the intended purpose of the skill for sharing components with a team or the public registry.
  • [CREDENTIALS_UNSAFE]: The skill describes authentication mechanisms involving API keys (21st_sk_...) and the storage of session tokens in ~/.config/21st/auth.json. It also suggests passing these keys via command-line flags or environment variables.
  • [PROMPT_INJECTION]: The skill processes component code and metadata retrieved from the 21st.dev registry, which represents a surface for indirect prompt injection if the source content is malicious. However, no direct injection attempts were found in the skill itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 5, 2026, 10:11 PM
Security Audit — agent-trust-hub — 21st-cli