The Agent Skills Directory

[PROMPT_INJECTION]: The skill follows standard instructional patterns for role-play as an 'art director'. No bypasses, safety filter overrides, or malicious system prompt extraction techniques were identified.
[DATA_EXFILTRATION]: No network operations (curl, wget, fetch) or sensitive file access patterns are present. The skill operates entirely within the agent's conversational context.
[COMMAND_EXECUTION]: There are no subprocess calls, shell commands, or platform-specific execution patterns (!command) that would indicate unauthorized command execution.
[EXTERNAL_DOWNLOADS]: The skill does not download external scripts, packages, or code. It references other internal skills (e.g., neuromarketing-ai) using standard link syntax, which is handled by the platform's orchestration layer.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user input via the AskUserQuestion tool to generate a Technical Assignment (TZ). While it lacks explicit boundary markers to wrap untrusted input, its capabilities are limited to text generation with no access to dangerous tools (file-write, network, or code execution), making the risk negligible.
[OBFUSCATION]: Analysis of the text and instructions revealed no hidden layers, Base64 strings, zero-width characters, or homoglyph substitutions.

brief-interview