The Agent Skills Directory

[SAFE]: The skill provides architectural and operational guidelines aimed at improving code quality and maintaining framework consistency. No malicious patterns, obfuscation, or unauthorized data exfiltration attempts were detected.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface and implements a defensive protocol.
Ingestion points: The skill notes that untrusted inputs may enter the system via BrowserOS, web content, or other external data sources.
Boundary markers: The provided instructions do not explicitly define technical boundary markers like XML tags, but they do define conceptual boundaries for operational modes (Build, Incident, Experiment).
Capability inventory: The framework allows for terminal execution via ExecutionProxy, file operations via MemoryStore (Obsidian docs), and web navigation via BrowserOS.
Sanitization: The instructions mandate a manual sanitization step, requiring the agent to redact secrets and PII before sharing external data with subagents.
[COMMAND_EXECUTION]: The protocols describe an ExecutionProxy Interface (using the rtk prefix) intended to manage and track terminal actions. This is documented as a standard operational procedure for the framework to ensure deterministic execution and token efficiency.

super-architect