The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides the agent with access to user emails, which are an untrusted external data source.
Ingestion points: Email subjects and bodies are retrieved through the Microsoft Graph API in scripts/microsoft.ts using the listMessages function.
Boundary markers: The skill does not implement any specific markers or instructions to isolate email content from the agent's primary instructions.
Capability inventory: The agent can read user profile data, download emails as .eml files to the local system, and communicate with Microsoft's servers.
Sanitization: While filename sanitization is performed to prevent file system issues, the email content is passed to the agent without additional sanitization or filtering.
[COMMAND_EXECUTION]: The scripts/lib/auth.ts script executes local commands to facilitate the OAuth flow and credential management.
It spawns system processes to open the user's web browser for login and uses the 1Password CLI (op) if available to retrieve API keys from a local vault.
These operations use fixed command strings or verified URLs and do not expose the system to arbitrary command injection from untrusted external sources.

microsoft-outlook