The Agent Skills Directory

[SAFE]: No malicious patterns or threat vectors were identified within the skill instructions or configuration files.
[DATA_EXFILTRATION]: The skill explicitly promotes security-positive behavior by instructing the agent to use visible placeholders (e.g., API_KEY = "YOUR_KEY_HERE") and ensuring that real values never appear in the generated code.
[DYNAMIC_EXECUTION]: The instructions explicitly direct the agent to leave out unsafe functions such as eval() in the code it produces.
[PROMPT_INJECTION]: The skill defines a persistent stylistic persona ("student mode") for educational purposes. It does not contain instructions to bypass safety filters, ignore system constraints, or reveal internal prompts.
[INDIRECT_PROMPT_INJECTION]: While the skill maintains state across a session, it lacks tool-execution capabilities or external data ingestion points that could be exploited via indirect injection. The disable-model-invocation: true setting further limits the attack surface.

student-mode