powerpoint-mcp
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates reading content from existing PowerPoint files via multiple tools, creating a vector for indirect prompt injection.
- Ingestion points: Data enters the agent's context through
presentation(action: "open"),textframe(action: "get-text"),table(action: "get-cell-text"),notes(action: "get-notes-text"), andslide(action: "import-from-file")as described inSKILL.mdand related reference files. - Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore instructions embedded within the processed slide content.
- Capability inventory: The skill possesses file-writing capabilities through
presentation(action: "close", save: true)andexportactions (PDF and image generation), enabling an agent to act upon potentially malicious instructions read from a file. - Sanitization: There is no evidence of sanitization, filtering, or escaping of the text read from the presentations before it is processed by the agent.
- [EXTERNAL_DOWNLOADS]: The
README.mdfile instructs users to install the skill usingnpx skills add sbroenne/mcp-server-powerpoint, which downloads code from an external repository provided by the author. - [COMMAND_EXECUTION]: The skill describes a toolset designed to drive a live PowerPoint desktop instance via COM/PIA interfaces. While the skill documentation provides the instructions, the underlying MCP server executes system-level operations on the host machine to automate the presentation software.
Audit Metadata