The Agent Skills Directory

[SAFE]: The skill is designed with a robust 10-stage workflow that prioritizes research integrity. It includes mandatory checkpoints that require explicit user confirmation before proceeding to subsequent stages, ensuring human-in-the-loop control.
[SAFE]: Mandatory integrity gates (Stage 2.5 and Stage 4.5) are implemented to perform 100% verification of citations, references, and quantitative claims. This is a best-practice defense against AI-generated hallucinations.
[COMMAND_EXECUTION]: The skill dispatches document formatting tasks to industry-standard tools like Tectonic (for LaTeX-to-PDF compilation) and Pandoc (for DOCX conversion). These tools are used for their intended purpose in an academic context.
[EXTERNAL_DOWNLOADS]: Verification processes involve querying the Semantic Scholar API and other established academic databases (e.g., Google Scholar, DOI resolvers). These are well-known, trusted services for bibliographic verification.
[PROMPT_INJECTION]: The orchestrator and its sub-agents contain explicit instructions to avoid 'sycophantic concession' and to maintain quality even when user input might push for shortcuts. It includes an 'Anti-Patterns' section that forbids skipping mandatory safety and integrity checks.

academic-pipeline