The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/intake_wizard.py executes os.system("chcp 65001 > nul") to configure the Windows console to use UTF-8 encoding. This is a standard procedure for ensuring correct character rendering in terminal interfaces.
[PRIVILEGE_ESCALATION]: The skill instructions in SKILL.md and the launcher scripts/launch_paperspine_ui.ps1 utilize require_escalated sandbox permissions and the PowerShell -ExecutionPolicy Bypass flag. These mechanisms are employed to launch a new terminal window for the interactive configuration wizard, which is the primary user interface for setting up project parameters.
[EXTERNAL_DOWNLOADS]: The citation verification script scripts/citation_verification_zh.py makes outbound HTTPS requests to api.crossref.org using urlopen. This activity is used to validate Digital Object Identifiers (DOIs) for academic references by fetching scholarly metadata.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data such as manuscripts and research materials, with ingestion points located in scripts/material_inventory.py and scripts/structured_review.py. While the skill employs TeX/Markdown stripping and normalization as sanitization in scripts/_paper_spine_utils.py, and SKILL.md designates user materials as authoritative (boundary markers), the combination of external data ingestion with local shell execution capabilities (the capability inventory including os.system and Start-Process) represents a potential surface for indirect prompt injection.
[DYNAMIC_EXECUTION]: The skill dynamically assembles and executes shell commands to launch its internal Python scripts through PowerShell and Bash wrappers (scripts/launch_paperspine_ui.ps1, scripts/launch_paperspine_ui.sh). These operations are used for UI orchestration and include basic string escaping to handle local paths and arguments.

paper-spine