The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references optional external dependencies, including search-cli from a third-party Homebrew tap (199-biotechnologies/tap) and the weasyprint Python library for generating PDF reports.
[COMMAND_EXECUTION]: The orchestration engine and quality control pipeline involve executing several local Python scripts (research_engine.py, validate_report.py, verify_citations.py, md_to_html.py) and external CLI tools for multi-source information retrieval.
[DATA_EXFILTRATION]: The verify_citations.py script performs outbound network requests to confirm the existence of DOIs and the accessibility of URLs cited in the research bibliography.
[PROMPT_INJECTION]: The skill operates as an autonomous research agent, processing external data from web sources which presents an indirect prompt injection surface.
Ingestion points: Untrusted data enters the agent's context through search results and web page content fetched via WebSearch, WebFetch, and search-cli (documented in reference/methodology.md).
Boundary markers: The skill requests structured evidence objects from its sub-agents to improve data integrity.
Capability inventory: The skill possesses capabilities for filesystem writes (Write/Edit tools) and local script execution.
Sanitization: The pipeline includes dedicated scripts for source credibility evaluation (source_evaluator.py), structural validation (validate_report.py), and citation hallucination detection (verify_citations.py).

deep-research