The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to run curl and jq for API interactions. The commands use fixed templates that prevent arbitrary command injection.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external data from emails and calendar events.
Ingestion points: Data enters the context via curl requests to the Microsoft Graph API in SKILL.md.
Boundary markers: No specific delimiters are used to isolate untrusted email content from the agent's system instructions.
Capability inventory: The skill uses the Bash tool, which has the capability to perform network requests and write files.
Sanitization: Content from email and event bodies is processed without sanitization or filtering.
Mitigation: The skill enforces a 'prepare → present → execute' workflow, which ensures that no destructive or outgoing action is taken without explicit user consent.

microsoft-outlook