The Agent Skills Directory

[COMMAND_EXECUTION]: The intake_wizard.py script executes shell commands through os.system() to clear the terminal screen and configure Windows console encoding (chcp 65001).
[PRIVILEGE_ESCALATION]: The skill instructions direct the agent to launch PowerShell using the -ExecutionPolicy Bypass flag, which intentionally circumvents local script execution restrictions to run the intake UI.
[COMMAND_EXECUTION]: The launch_paperspine_ui.sh script uses osascript on macOS and various Linux terminal emulators (e.g., gnome-terminal, xfce4-terminal) to spawn new interactive processes for the wizard UI.
[INDIRECT_PROMPT_INJECTION]: The skill ingests text from user-provided drafts and materials to automatically infer project configuration, which is a vector for indirect prompt injection if those files contain malicious instructions.
Ingestion points: scripts/humanize_check.py reads manuscript files in final_paper/; scripts/intake_wizard.py recursively reads text and markdown files in the project directory.
Boundary markers: None identified; content is read directly into variables and used to populate configuration fields.
Capability inventory: The skill has the ability to execute shell commands and launch external terminal windows.
Sanitization: No sanitization or filtering is applied to the ingested text before it is processed by the agent.

paper-spine-intake