The Agent Skills Directory

[SAFE]: The skill demonstrates high-quality design with strict role boundaries and data management protocols. All file operations are restricted to the local ./brand-brain/ workspace.
[DYNAMIC_EXECUTION]: The skill includes and executes a local Python utility (_system/scripts/memory_manager.py) for processing structured memory logs. Analysis confirms the script performs benign operations such as JSON parsing, data filtering, and time-based decay calculation without network access or unsafe evaluation.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-generated learnings stored in a JSONL file. While this represents a potential attack surface, the system implements robust defensive instructions (e.g., in _system/preamble.md and _system/edge-cases.md) to validate and filter these memories before use.
[DATA_EXPOSURE]: No hardcoded credentials, sensitive file path access, or unauthorized network exfiltration patterns were detected. Data handling is limited to the business context provided by the user.
[PROMPT_INJECTION]: The skill includes specific defensive instructions in _system/iron-rules.md and _system/edge-cases.md to prevent the leakage of internal system prompts and protocols, maintaining a secure operational boundary.

afa