lov-event-curator
Warn
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: MEDIUMOBFUSCATIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [OBFUSCATION]: The core skill instructions are provided in an encrypted format (
SKILL.md.enc) utilizing AES-256-GCM as specified inMANIFEST.enc.json. This implementation conceals the actual logic and prompts from static security review until decrypted at runtime. - [REMOTE_CODE_EXECUTION]: Setup involves running
npx lovstudioanduvx lovstudio-skill-helper, which execute code from vendor-maintained packages on NPM and PyPI to facilitate installation and decryption. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted 'guest background material' to generate plans, creating a potential vulnerability to embedded malicious instructions. Ingestion points: Guest background material data (mentioned in SKILL.md). Boundary markers: None identified in metadata; the body is encrypted. Capability inventory: Event planning and text generation. Sanitization: None verifiable due to encrypted payload.
- [EXTERNAL_DOWNLOADS]: The skill requires downloading external CLI tools from public registries (NPM and PyPI) using standard package runners.
- [COMMAND_EXECUTION]: The skill instructions include shell commands for adding and decrypting the skill components.
Audit Metadata