The Agent Skills Directory

Prompt Injection (SAFE): No malicious prompt injection patterns were detected in the instruction set. Standard instructional language is used for component integration workflows.\n- Data Exposure & Exfiltration (SAFE): No evidence of sensitive data access or unauthorized network communication was found. File access in scripts is limited to public project configuration files like package.json.\n- Obfuscation (SAFE): No encoded or hidden content was found in the skill files.\n- Unverifiable Dependencies & Remote Code Execution (SAFE): The skill uses official CLIs and package runners from trusted sources like shadcn and Google Labs. No piping of untrusted remote scripts to the shell was detected.\n- Privilege Escalation (SAFE): No commands for acquiring administrative privileges or modifying system-level settings were found.\n- Persistence Mechanisms (SAFE): No attempts to maintain access across sessions via shell profiles or scheduled tasks were identified.\n- Metadata Poisoning (SAFE): Metadata fields accurately describe the skill's purpose without deceptive content.\n- Indirect Prompt Injection (SAFE): The skill ingests component data from external registries, which is the intended use case. Documented surface includes ingestion via MCP tools, presence of write capabilities, and lack of sanitization logic, but no active exploitation markers were found.\n- Time-Delayed / Conditional Attacks (SAFE): No logic gating malicious behavior based on dates or environmental triggers was found.\n- Dynamic Execution (SAFE): No runtime code generation or unsafe deserialization of external data was identified.

shadcn-ui