The Agent Skills Directory

Command Execution & Dynamic Scripting: The skill provides specific Python code snippets for the agent to manage file operations, including copying binaries to workspace directories and safely removing slides from PowerPoint presentations. These instructions are intended for the skill's primary function and include procedures to prevent file corruption.
Indirect Prompt Injection Surface: The skill ingests and processes external Word and PowerPoint templates provided via SharePoint or local storage. Since these files enter the agent's context, they represent a potential surface for indirect prompt injection. The skill mitigates this by instructing the agent to ignore embedded macros and validate that the final output adheres to brand rules without inventing new content.
Ingestion points: Templates are retrieved from assets/ or SharePoint knowledge sources configured in assets/template-manifest.json.
Boundary markers: The instructions specify that the agent should preserve user meaning while only replacing conflicting terminology and must not invent brand claims.
Capability inventory: The skill uses Python script execution (via shutil and pptx library patterns), file writing, and the built-in analyzing-pptx tool.
Sanitization: Instructions require the agent to check for leftover placeholder text and explicitly forbid running macros or embedded executables.
External Resource Retrieval: The skill fetches Office binaries from user-defined SharePoint URLs. This functionality is a core requirement for template management and utilizes the user's configured SharePoint infrastructure and agent knowledge permissions.
File System Interaction: The instructions guide the agent to move files from temporary upload paths (/app/uploads/) to workspace paths (/app/workspace/). This is a standard practice to ensure compatibility with downstream analysis tools like analyzing-pptx.

brand-template-enforcer