The Agent Skills Directory

[DATA_EXFILTRATION]: The skill communicates with graph.microsoft.com and login.microsoftonline.com to retrieve directory configuration and audit logs. These are well-known Microsoft services, and the operations are consistent with the skill's stated purpose of auditing cloud security. No unauthorized or suspicious network destinations were detected.- [COMMAND_EXECUTION]: The provided Python script does not utilize any subprocess calls or shell execution commands.- [REMOTE_CODE_EXECUTION]: No remote script downloads or dynamic code execution patterns (such as eval or exec) were identified in the analysis.- [CREDENTIALS_UNSAFE]: The skill handles authentication secrets through standard command-line arguments. No hardcoded credentials or insecure secret storage practices were observed.- [PROMPT_INJECTION]: The instructions in the skill body do not contain any patterns attempting to override agent behavior, bypass safety filters, or extract system prompts.

detecting-suspicious-oauth-application-consent