The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documents the installation of parallel-cli by fetching a script from parallel.ai. This is a well-known service providing the academic search tools used throughout the skill's workflow.
[COMMAND_EXECUTION]: The script scripts/generate_pdf.py uses the subprocess module to invoke system-level tools such as pandoc and xelatex for converting markdown content into professionally formatted PDF documents.
[COMMAND_EXECUTION]: The script scripts/generate_schematic.py executes internal Python components via subprocess to coordinate AI-powered diagram generation and refinement.
[INDIRECT_PROMPT_INJECTION]: The skill processes research data, abstracts, and citations from external academic databases, which constitutes a surface for potential prompt injection if malicious content is embedded in the processed literature.
Ingestion points: scripts/verify_citations.py (markdown file), scripts/search_databases.py (JSON search results), and scripts/generate_pdf.py (markdown document).
Boundary markers: Relies on structured templates and script-based parsing but does not include explicit 'ignore instructions' delimiters for all dynamic content.
Capability inventory: Subprocess execution in scripts/generate_pdf.py and network operations in scripts/verify_citations.py for metadata retrieval.
Sanitization: Implements regex-based filtering for DOI extraction and uses structured JSON handling to mitigate risks from malformed inputs.
[DYNAMIC_EXECUTION]: The scripts/generate_schematic_ai.py script dynamically constructs prompts sent to an external AI model (OpenRouter) to generate and critique scientific schematics based on user descriptions.

literature-review