The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data, including Teams AI meeting transcripts, emails, and chat messages. This represents a potential indirect prompt injection surface. However, the risk is mitigated by explicit quality gate instructions in convert-meeting-minutes.prompt.md and references/material-lifecycle.md that command the agent to omit credentials, token-bearing URLs, and sensitive commercial terms from generated records. The skill's primary capabilities are limited to filesystem organization and text summarization.
[COMMAND_EXECUTION]: The skill provides two PowerShell scripts: Initialize-CustomerWorkspace.ps1 and Test-ReceivedMaterialPlacement.ps1. The initialization script automates the creation of a standard directory structure and copies templates from the skill's assets. The audit script performs local file signature checks (PDF, ZIP, OLE) to verify file formats. Both scripts operate locally, do not fetch remote payloads, and perform standard filesystem tasks consistent with the skill's stated purpose.

customer-workspace