The Agent Skills Directory

[DYNAMIC_EXECUTION]: Runtime compilation of C source code and process injection.\n
Script: scripts/office/soffice.py\n
Evidence: Contains a hardcoded C source code string (_SHIM_SOURCE) that is written to a temporary file and compiled into a shared object using gcc at runtime.\n
Evidence: Employs the LD_PRELOAD environment variable to inject the resulting library into the soffice process. This is designed to intercept and modify low-level system calls (socket, listen, accept) to bypass system-enforced restrictions on UNIX sockets.\n- [PERSISTENCE]: Modification of local application configuration.\n
Script: scripts/recalc.py\n
Evidence: Automatically installs a StarBasic macro (RecalculateAndSave) into the user's LibreOffice configuration directory (e.g., ~/.config/libreoffice/). This configuration persists across sessions and modifies the behavior of the local Office installation.\n- [COMMAND_EXECUTION]: Execution of system compilers and application binaries.\n
Script: scripts/office/soffice.py, scripts/recalc.py, scripts/office/validators/redlining.py\n
Evidence: Uses subprocess.run to execute gcc, soffice, git, and timeout utilities.\n- [INDIRECT_PROMPT_INJECTION]: Significant attack surface due to document processing capabilities.\n
Ingestion points: scripts/recalc.py and scripts/office/unpack.py process external .xlsx and .docx files.\n
Boundary markers: Absent. There are no instructions to ignore embedded commands within the processed data.\n
Capability inventory: High-risk capabilities including arbitrary file writes and shell command execution via subprocess.\n
Sanitization: While the skill correctly uses defusedxml to prevent XXE attacks, it does not sanitize content or file paths before they are used in shell commands.

xlsx