The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and analyze user-provided manuscript chapters and outlines to update its internal memory bank. This ingestion of untrusted data constitutes a typical indirect prompt injection surface. However, the skill explicitly mitigates this in its 'Content Boundary and Sanitization Rules' within references/book_memory_protocol.md, which instructs the agent to treat all user-authored content as narrative data only, ignore any instruction-like text encountered in files, and sanitize memory updates by restructuring information into factual summaries.
[COMMAND_EXECUTION]: The skill provides automated manuscript compilation workflows. It utilizes local shell scripts (combine_chapters.sh, combine_chapters.ps1) and batch files (generate_docx.bat) to aggregate chapter files and convert them to Word documents using Pandoc. These scripts are transparent and restricted to managing files within the user's project workspace.
[EXTERNAL_DOWNLOADS]: The initialization workflow suggests the installation of companion skills (research-dojo and calligraph-sensei) via npx skills add. These dependencies originate from the same author's GitHub repository (github.com/kshanxs/skill-dojo), representing a standard extension of the skill's environment using platform-supported tools.

book-writer