outlook-api
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides the ability to ingest untrusted data from external sources, which constitutes an attack surface for indirect prompt injection.
- Ingestion points: The skill reads external content from Microsoft Outlook messages (
/messages) and calendar events (/events), which are typically authored by third parties. - Boundary markers: No specific boundary markers or "ignore embedded instructions" warnings are included to delimit the untrusted data from the agent's instructions.
- Capability inventory: The skill possesses write capabilities including sending emails (
/sendMail), creating calendar events, and managing contacts, which could be triggered by malicious instructions embedded in incoming mail. - Sanitization: There is no evidence of content sanitization or validation for the data retrieved from the API.
- [COMMAND_EXECUTION]: The skill documentation instructs the agent to perform tasks by generating and executing Python scripts using shell heredocs.
- Evidence: Numerous examples in
SKILL.md(e.g., in the 'Quick Start' and 'Connection Management' sections) utilize thepython <<'EOF'pattern to execute code that makes network requests and processes environment variables.
Audit Metadata