The Agent Skills Directory

[COMMAND_EXECUTION]: Runtime compilation and process injection are used to shim system calls.
Evidence: scripts/office/soffice.py writes C source code to a temporary file and compiles it into a shared library using gcc.
Evidence: It executes soffice with LD_PRELOAD set to the path of the compiled shared object to facilitate socket operations in restricted environments.
[COMMAND_EXECUTION]: Several scripts invoke system-level utilities.
Evidence: scripts/recalc.py and scripts/office/validate.py perform subprocess calls to soffice, git, and gtimeout.
[PROMPT_INJECTION]: The skill processes untrusted spreadsheet data and possesses powerful tool capabilities, creating a surface for indirect prompt injection.
Ingestion points: Data from spreadsheet files is ingested using pandas and openpyxl.
Boundary markers: No delimiters or instructions are used to isolate untrusted data from the agent prompt.
Capability inventory: The skill can execute shell commands and write to the local file system.
Sanitization: No sanitization or validation of the ingested data is performed.

xlsx