The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled Node.js script (scripts/dossier_to_docx.js) to convert Markdown research dossiers into Word documents. This involves standard command-line invocation using Node.js to process local project files.
[EXTERNAL_DOWNLOADS]: The skill relies on the docx npm package, a well-known library for generating Word documents. Documentation and instructions are provided for installing this dependency through the standard npm registry.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external research databases via search tools.
Ingestion points: Paper metadata, including titles, abstracts, and identifiers, are retrieved using the research-hub search tool and processed into the topic_dossier.md file.
Boundary markers: The skill expects structured JSON output from its search tools, which helps delineate external content from agent instructions.
Capability inventory: The agent has the ability to write and read files in the .research/ directory and execute the local dossier_to_docx.js script.
Sanitization: The skill implements "Honesty rules" requiring verbatim quote verification and the resolution of academic identifiers (DOIs/arXiv IDs) to ensure the integrity of the evidence presented in the dossier.
[DYNAMIC_EXECUTION]: The workflow involves the execution of a provided Node.js script that uses the fs and path modules to perform file transformations. The script's logic is transparent and limited to reading Markdown content and generating a formatted .docx file.

gap-to-topic