The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or data exfiltration techniques were detected across the skill files.
[SAFE]: The skill does not request or utilize any sensitive tools, network access, or filesystem permissions.
[SAFE]: The workflow is purely conversational, focusing on pedagogical probing and architectural critique without executing code or scripts.
[SAFE]: The examples provided in examples.md demonstrate best practices for identifying security flaws in user designs (e.g., plaintext passwords in JWTs) rather than introducing them into the agent's behavior.

architecture-review