The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data from academic databases (e.g., PubMed, Google Scholar, arXiv) and user-provided research topics. It incorporates robust defensive measures to mitigate indirect prompt injection risks. These include a systematic reference verification protocol in agents/source_verification_agent.md (DOI and metadata checks), an evidence grading hierarchy in references/source_quality_hierarchy.md, and mandatory AI disclosure statements in all outputs (agents/report_compiler_agent.md). The capability inventory is restricted to research documentation and synthesis tasks, with no execution of arbitrary code derived from external inputs.
[EXTERNAL_DOWNLOADS]: The skill references and encourages the use of established academic and research-oriented services for literature monitoring and preregistration. These include well-known platforms such as Google Scholar, PubMed, arXiv, bioRxiv, OSF, and PROSPERO. All referenced domains are official, well-known services within the scientific community and do not pose a security risk.

deep-research