The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The script contains a hardcoded absolute file system path that exposes the author's local username and folder structure.
Evidence: WORK_FOLDER_ROOT in scripts/contract_kickoff.py is set to C:\Users\AGasser\OneDrive\5 ANE CLAUDE work folder.
Context: While this facilitates loading the vendor's package (ane_package), it leaks environment-specific metadata.
[DYNAMIC_EXECUTION]: The skill uses a PowerShell script block to interact with Excel via COM objects for calculation verification.
Evidence: The _VERIFY_BODY string in scripts/contract_kickoff.py contains PowerShell code executed via excelcom.apply_com.
Context: The execution is used to verify workbook arithmetic. The script attempts to mitigate injection by casting input values to [double] and sanitizing role labels using a slugging function.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (Word documents) that influences generated output and automation logic.
Ingestion points: The read command in scripts/contract_kickoff.py ingests .docx files through load_document.
Boundary markers: No explicit boundary markers or "ignore embedded instructions" warnings are applied to the text extracted from the contracts.
Capability inventory: The skill can create directories, write Markdown/Word files, and execute PowerShell scripts for Excel automation.
Sanitization: Labels are sanitized via _slug (limiting characters to alphanumeric and dashes) and numbers are validated via parse_number, providing some protection against control character injection into the PowerShell logic.

contract-delivery-kickoff