The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script 'scripts/office/soffice.py' contains a C source code string that is written to a temporary file, compiled using 'gcc' into a shared library, and then deleted at runtime.
[PRIVILEGE_ESCALATION]: The file 'scripts/office/soffice.py' spawns the LibreOffice process using the 'LD_PRELOAD' environment variable to inject the locally compiled shared library. This technique is used to intercept and shim system calls like socket and accept, which is a high-risk process manipulation behavior.
[PERSISTENCE_MECHANISMS]: The script 'scripts/recalc.py' writes a StarBasic macro file named 'Module1.xba' to the LibreOffice application support directories in the user's home folder, which persists and affects the application's environment.
[INDIRECT_PROMPT_INJECTION]: 1. Ingestion points: Spreadsheet data is ingested through 'pandas.read_excel' and 'openpyxl.load_workbook' as described in 'SKILL.md'. 2. Boundary markers: Absent. The prompt instructions do not include delimiters or specific instructions to treat ingested tabular data as untrusted. 3. Capability inventory: The skill has the ability to execute subprocesses (gcc, soffice, git) and perform file system writes. 4. Sanitization: The skill uses 'defusedxml' for XML parsing, which provides some protection against standard XML external entity and expansion attacks.

xlsx