calendar-event-manager
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
Bashtool to execute AppleScript commands viaosascript. This allows the agent to interact directly with the macOS Calendar and Reminders applications, which could be misused if the generated scripts are not properly constrained. - [DYNAMIC_EXECUTION]: The skill dynamically generates AppleScript code at runtime by interpolating user-provided data (e.g., event names, locations, notes) into predefined code templates found in
SKILL.md. This pattern of script generation and execution presents a risk if the input contains control characters or malicious payloads. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, specifically calendar invitations and user-provided event descriptions, to automate schedule creation.
- Ingestion points: Event details extracted from invitations or user descriptions as defined in 'Template A' and 'Template B' in
SKILL.md. - Boundary markers: None identified; input data is processed directly without explicit delimiters or instructions to ignore embedded commands.
- Capability inventory: The skill can execute shell commands via
Bash, interact with local databases (Calendar/Reminders), and modify local files. - Sanitization: There are no instructions for sanitizing or escaping the input before it is used in AppleScript construction.
- [PERSISTENCE]: The skill includes a 'Self-Evolving' directive that explicitly instructs the agent to modify the
SKILL.mdfile to fix errors or update logic. While intended for maintenance, this self-modification capability could be exploited by an attacker to persist malicious instructions in the agent's primary skill set.
Audit Metadata