The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill suite contains multiple ingestion points for untrusted external data in the form of academic papers (PDFs and ArXiv source files).
Ingestion points: research-paper-kb/SKILL.md, slr-automation-guide/SKILL.md, and scientify-idea-generation/SKILL.md read and analyze contents from external papers.
Boundary markers: None identified. Instructions do not specify delimiters or warnings to ignore embedded instructions within the papers.
Capability inventory: scientify-idea-generation/SKILL.md utilizes shell commands including git clone, curl, and tar to manage research resources.
Sanitization: No explicit sanitization or filtering of the content extracted from academic papers is described before it is processed by the agent or interpolated into prompts.
[COMMAND_EXECUTION]: scientify-idea-generation/SKILL.md employs shell commands to manage the research workspace. While these are intended for legitimate research data acquisition, the use of git clone, curl, and tar with variable inputs (like repository URLs and ArXiv IDs) represents a standard command execution surface.
[EXTERNAL_DOWNLOADS]: The skills reference and download content from several well-known and trusted services.
Fetches paper source data from ArXiv (arxiv.org).
Clones methodology tools and guides from GitHub repositories including vitorfs/parsifal, asreview/asreview, SakanaAI/AI-Scientist, and poloclub/awesome-grad-school.

methodology-skills