The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it extracts and processes text content from external PowerPoint files which is then used to build prompts for the Gemini API.
Ingestion points: scripts/extractor.py extracts all text, chart data, and table content from user-supplied PPTX files.
Boundary markers: In scripts/mapper.py and scripts/pptx_converter.py, the extracted data is structured into JSON blocks within the prompt, but there are no explicit delimiters or system instructions to ignore potential commands embedded in the slide data.
Capability inventory: The skill has the capability to write files to the disk (scripts/generator.py using prs.save) and execute shell commands to run its internal logic as described in SKILL.md.
Sanitization: No evidence was found of content filtering or sanitization of the strings extracted from the PowerPoint files before they are interpolated into the LLM prompts.
[COMMAND_EXECUTION]: The skill documentation and functionality rely on the execution of Python scripts via the shell (e.g., python scripts/pptx_converter.py convert). While this is necessary for the skill's primary purpose of file manipulation, it establishes a pattern of command execution on the host environment.

pptx-converter