The Agent Skills Directory

[SAFE]: The skill is entirely composed of documentation and instructional markdown. It does not include executable code, scripts, or commands that could pose a security risk. All operations described are functional instructions for the agent to use specific tools.
[CREDENTIALS_SAFE]: The skill references an environment variable (ZOOM_MEETING_MCP_CREDENTIAL) for authentication. This is a standard and secure practice for secret management, and no hardcoded credentials or private keys were detected in the file.
[PROMPT_INJECTION]: The instructions provide functional guidelines for data handling and reporting (e.g., 'If nothing relevant comes back, say so plainly'). There are no attempts to bypass safety filters, act as an unrestricted agent (DAN), or override system instructions.
[INDIRECT_PROMPT_INJECTION]: Although the skill processes external data from meeting histories and network states, it defines a read-only correlation workflow. It does not include capabilities such as file writing, shell execution, or remote network requests that would facilitate an indirect injection attack. The instructions emphasize accuracy and evidence-based reporting.

zoom-meeting-context