The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/office/soffice.py contains a hardcoded C source string (_SHIM_SOURCE) which is written to a temporary file, compiled at runtime using gcc, and then injected into the LibreOffice (soffice) execution environment using the LD_PRELOAD environment variable. This technique is used to intercept and redirect system calls to allow the application to function in restricted or sandboxed environments.
[COMMAND_EXECUTION]: The skill frequently uses subprocess.run to execute external system utilities. These include gcc for runtime compilation in scripts/office/soffice.py, soffice and pdftoppm for document conversion in scripts/thumbnail.py, and git diff for comparing document changes in scripts/office/validators/redlining.py.
[EXTERNAL_DOWNLOADS]: The skill documentation in SKILL.md and pptxgenjs.md specifies dependencies on external packages from official registries.
Evidence: Instructions to pip install Pillow and npm install -g pptxgenjs.
[INDIRECT_PROMPT_INJECTION]: The skill provides the ability to extract text from user-supplied PowerPoint files, which is then processed by the agent. This ingestion of untrusted data creates a potential surface for indirect prompt injection.
Ingestion points: The extract-text command in SKILL.md extracts text from .pptx files provided as input.
Boundary markers: No explicit delimiters or warnings to ignore embedded instructions are used when passing extracted content to the agent.
Capability inventory: The skill has the capability to read/write to the file system and execute shell commands via subprocess.run across multiple scripts.
Sanitization: There is no evidence of filtering or sanitization of the text extracted from documents.

pptx