The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Runtime code generation and execution detected. The file scripts/office/soffice.py contains an embedded C source string that is written to a temporary file, compiled with gcc, and loaded into the process space via the LD_PRELOAD environment variable to handle socket restrictions. The file scripts/recalc.py also generates a StarBasic macro and writes it to the application configuration path to facilitate automated formula recalculation in LibreOffice.
[COMMAND_EXECUTION]: Systematic execution of system commands. The skill invokes utilities including gcc, git, soffice, and timeout via subprocess.run to perform its tasks. Evidence chain for indirect prompt injection surface: 1. Ingestion points: scripts/office/unpack.py (unpacks user-provided Office documents). 2. Boundary markers: Absent; instructions lack explicit delimiters for document data. 3. Capability inventory: scripts/office/soffice.py (compilation), scripts/recalc.py (soffice automation), scripts/office/validators/redlining.py (git integration). 4. Sanitization: scripts/office/security_utils.py implements safe XML parsing with defusedxml and provides path validation utilities.

xlsx