The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/officecli_xml_common.py uses subprocess.run to execute the officecli binary. While the arguments are passed as a list (preventing shell injection), the specific operations (XPaths, XML payloads) are derived from user-provided JSON specification files. This is the intended primary purpose of the skill for standards document orchestration.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection as it processes external JSON specifications and Word document content to perform modifications.
Ingestion points: The scripts add_bookmarks.py, fix_enum_tabs.py, fix_heading_tabs.py, and fix_note_tabs.py ingest untrusted data from JSON files passed via command-line arguments.
Boundary markers: No explicit delimiters or boundary markers are implemented in the Python logic to separate trusted instructions from data, though the skill provides JSON schemas to define expected structures.
Capability inventory: The skill has the capability to read local files, execute subprocesses (via officecli_xml_common.py), and modify document XML.
Sanitization: Input is parsed via json.load. While the skill includes JSON schemas for validation in assets/schemas/, the provided Python scripts do not explicitly execute schema validation before processing.

sdo-docx-operations