The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external data which serves as a potential vector for indirect instructions.\n
Ingestion points: The skill reads DOCUMENT-ACCESSIBILITY-AUDIT.md and processes the contents of .docx, .xlsx, and .pptx files to determine required fixes.\n
Boundary markers: No explicit boundary markers or directives to ignore embedded instructions are present in the processing logic.\n
Capability inventory: The skill has the capability to write new document files, execute Python scripts, and install third-party packages.\n
Sanitization: There is no evidence of content sanitization for data extracted from the documents before it is used in the remediation process.\n- [DYNAMIC_EXECUTION]: The skill generates and executes Python code at runtime to modify document structures.\n
Evidence: The remediation process includes generating Python scripts using python-docx, openpyxl, and python-pptx and executing them after user review.\n
Context: This is the primary functional mechanism of the skill, mitigated by rules requiring document backups and the use of specific file suffixes for output.\n- [COMMAND_EXECUTION]: The skill involves executing shell commands for package management and script execution.\n
Evidence: Rule 3 specifies that the agent must ask before installing required Python packages, implying the use of pip or similar tools.

office-remediator