The Agent Skills Directory

[COMMAND_EXECUTION]: The workspace provides several Python scripts, such as setup_workspace.py and build_review_pdf.py, that execute shell commands and Docker containers using the subprocess module. These commands are necessary for creating the project structure and running the Re:VIEW build toolchain.
[EXTERNAL_DOWNLOADS]: The skill refers to and utilizes external resources including the vvakame/review Docker image and the unicode-display_width Ruby gem. These are well-known and standard dependencies within the Re:VIEW publishing ecosystem.
[PROMPT_INJECTION]: The skill configures AI agents (e.g., @writing-reviewer) to read and process manuscripts in the 02_contents/ directory. This creates a surface for indirect prompt injection, as malicious instructions within a manuscript could potentially override the agent's review guidelines.
Ingestion points: 02_contents/ (Markdown manuscript files)
Boundary markers: None identified; manuscripts are processed directly as raw text.
Capability inventory: Agent permissions include reading and editing workspace files and running terminal commands for conversion and building.
Sanitization: No explicit sanitization of manuscript content before processing is present in the conversion scripts.

book-writing-workspace