pptx-design-agent
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes text from existing .pptx files, which could contain malicious instructions designed to influence the agent's behavior.\n
- Ingestion points: The skill includes logic to enumerate and read text from all shapes and speaker notes in a presentation (referenced in SKILL.md and applescript-patterns.md).\n
- Boundary markers: No delimiters or "ignore embedded instructions" warnings are used when the agent processes the ingested text.\n
- Capability inventory: The agent can execute arbitrary Python scripts (via python3) and AppleScript commands (via osascript), and has file system write access.\n
- Sanitization: No sanitization or filtering of the ingested slide content is performed before it is used to plan or execute modifications.\n- [EXTERNAL_DOWNLOADS]: The skill installs the 'python-pptx' and 'lxml' Python libraries from the standard package registry (PyPI).\n- [COMMAND_EXECUTION]: The skill performs shell command execution to automate PowerPoint. It uses 'python3' to run slide-building scripts and 'osascript' to interact with the PowerPoint application via AppleScript IPC.\n- [DYNAMIC_EXECUTION]: The skill generates Python and AppleScript code at runtime based on the user's request and templates provided in the reference documentation.
Audit Metadata