The Agent Skills Directory

Prompt Injection (SAFE): Instructions are strictly technical and do not contain patterns aimed at overriding agent constraints or bypassing safety filters.\n- Data Exposure & Exfiltration (SAFE): No network operations or access to sensitive system paths (e.g., credentials or configuration files) were detected. The asset validation script targets only the project's public directory to check file sizes.\n- Obfuscation (SAFE): Documentation and code are clear and do not utilize encoding, zero-width characters, or homoglyphs to mask intent.\n- Unverifiable Dependencies & Remote Code Execution (SAFE): The skill references standard packages such as zod and the Three.js ecosystem. No remote code execution or unsafe download patterns (e.g., curl|bash) were found.\n- Privilege Escalation (SAFE): No commands requiring sudo, administrative rights, or modification of system-level permissions are present.\n- Persistence Mechanisms (SAFE): No scripts or instructions for maintaining persistence across sessions or modifying shell profiles were identified.\n- Metadata Poisoning (SAFE): All metadata fields, including the description and author information, are professional and consistent with the skill's stated purpose.\n- Indirect Prompt Injection (SAFE): The asset scanning logic in scripts/validate-assets.ts only evaluates file metadata (extension and size) and does not ingest file content into the agent context, eliminating this attack vector.\n- Time-Delayed / Conditional Attacks (SAFE): No logic was found that gates functionality based on time-based triggers or specific environmental conditions.\n- Dynamic Execution (SAFE): The skill does not use runtime code generation (eval/exec) or unsafe deserialization of untrusted data.

threejs-expert