The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install msftrecon and o365spray via pip. These packages are hosted in community repositories and are not provided by trusted vendors. It also includes go install commands for tools from the projectdiscovery GitHub organization.
[COMMAND_EXECUTION]: The skill contains multiple shell command blocks for environment setup and tool invocation. It provides Python and Playwright script templates for automating credential sprays and user enumeration. A critical security weakness in the Python template is the code that disables SSL certificate and hostname validation (ssl.CERT_NONE, check_hostname=False), which exposes all authentication data to man-in-the-middle interception.
[INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by processing external data from network authentication responses and the local o365_attempts.json file. These ingestion points lack boundary markers or sanitization logic. The skill has capabilities to perform network requests (urllib.request.urlopen) and file system modifications (os.replace, json.dump), allowing processed external data to influence these privileged operations.

m365-entra-attack