The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the installation and execution of various third-party command-line security tools including msftrecon, o365spray, and the ProjectDiscovery toolkit (pd toolkit).
[EXTERNAL_DOWNLOADS]: Instructions include downloading security tools from external sources, specifically msftrecon and o365spray via pip, and ProjectDiscovery tools via go install from GitHub repositories.
[REMOTE_CODE_EXECUTION]: The skill provides Python scripts using urllib.request and playwright to perform automated network-based credential validation. The code intentionally disables SSL certificate verification (ssl.CERT_NONE and --ignore-certificate-errors) to allow the traffic to be intercepted by security research tools like Burp Suite, which introduces a Man-in-the-Middle (MitM) risk if used in untrusted environments.
[DATA_EXFILTRATION]: The validation scripts transmit user-supplied credentials (email and password) to Microsoft's authentication endpoints (login.microsoftonline.com). This is the primary intended function of the skill for credential verification.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes untrusted user-supplied data (email lists and credentials) through its validation logic.
Ingestion points: Untrusted data enters the agent context via the email and password parameters in the attempt() and saml_validate() functions.
Boundary markers: Absent; the scripts do not employ specific delimiters to separate user data from instruction logic.
Capability inventory: The skill possesses the capability to perform network requests (urllib, playwright), write to the local file system (json.dump), and execute shell commands (msftrecon).
Sanitization: The script utilizes urllib.parse.urlencode to sanitize inputs before transmission in POST requests.
[COMMAND_EXECUTION]: The skill recommends using the --break-system-packages flag with pip, which allows modification of the system-wide Python environment. This is a standard procedure for dedicated security testing environments but can impact system integrity.

m365-entra-attack