The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several internal Python scripts (e.g., research_engine.py, validate_report.py, verify_citations.py) to manage the research workflow and ensure report quality. These scripts are self-contained, utilize the Python standard library, and do not perform arbitrary or unsafe command execution.
[EXTERNAL_DOWNLOADS]: The verify_citations.py script makes network requests to doi.org and external URLs extracted from the research bibliography. These requests are used to resolve DOIs and verify the accessibility of sources, which is a core feature of the skill's anti-hallucination and verification protocol.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from the web through search tools. While this creates a surface for indirect prompt injection, the skill includes significant mitigation strategies, including a dedicated triangulation phase that requires cross-referencing claims across three or more independent sources and an automated validation gate.
[DATA_EXPOSURE]: The skill saves research findings and final reports to local directories such as ~/Documents/ and ~/.claude/research_output/. This is standard behavior for a document generation tool and is clearly documented within the skill's instructions.

deep-research