The Agent Skills Directory

[Prompt Injection] (SAFE): The instructions focus on grounding responses in financial theory and do not attempt to bypass AI safety guidelines or override system instructions.
[Data Exposure & Exfiltration] (SAFE): No hardcoded secrets, sensitive file paths, or network exfiltration patterns were identified.
[Obfuscation] (SAFE): All content is provided in human-readable markdown with no hidden characters, encoding, or obfuscated logic.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill does not include any scripts or package requirements (like requirements.txt or package.json) and contains no commands to download or execute remote code.
[Indirect Prompt Injection] (SAFE): While the skill defines patterns to analyze user-provided financial code, it lacks capabilities such as file-writing or network access that could be leveraged by malicious input.
[Privilege Escalation & Persistence] (SAFE): No commands are present that would modify system permissions, environment configurations, or startup routines.

risk-modeling