The Agent Skills Directory

[COMMAND_EXECUTION]: The pipeline utilizes the Bash tool to execute internal Python scripts for draft management, revision application, and integrity verification. It also automates the invocation of academic tools such as Pandoc and Tectonic to generate PDF and DOCX outputs from Markdown and LaTeX sources.- [EXTERNAL_DOWNLOADS]: The skill connects to academic services like Retraction Watch and various journal policy pages to verify citation integrity and retrieve submission guidelines. These operations target well-known academic resources and are used for scholarly auditing.- [PROMPT_INJECTION]: The skill features an extensive ingestion surface for external data, including academic search results and user-provided reviewer comments, which represents a surface for potential indirect prompt injection.
Ingestion points: literature_strategist_agent.md (academic database results) and revision_coach_agent.md (unstructured reviewer comments).
Boundary markers: While internal delimiters and HTML comment markers are used for document structure and audit trails, the skill lacks explicit instructions for agents to disregard commands that might be embedded in ingested academic text.
Capability: The agent has permissions for Bash, Write, and Edit tools, which could theoretically be abused if an indirect injection were successful.
Sanitization: Textual content from external sources is checked for writing quality and style but not for safety-related injection patterns.

academic-paper