The Agent Skills Directory

[SAFE]: The skill employs an orchestration pattern using sub-agents (researcher, evaluator, synthesizer) with restricted permissions (e.g., :read-only or :workspace) and disabled network access (web_search = "disabled"), following the principle of least privilege.- [SAFE]: A local Python validation script (scripts/validate_research.py) is used to ensure the integrity of research results. The script performs static checks (validating file existence and line counts) without executing content from the analyzed files.- [SAFE]: The orchestration instructions include explicit security boundaries, directing the main agent to avoid reading the content of generated research files to prevent potential context contamination or indirect prompt injection.- [SAFE]: No obfuscation, hardcoded credentials, or unauthorized network operations were identified across the skill instructions, agent configurations, or supporting scripts.

deep-research