The Agent Skills Directory

[SAFE] (SAFE): Technical review of the Python scripts and markdown documentation confirms the content is educational and serves as a framework for building multi-agent systems.
[INDIRECT_PROMPT_INJECTION] (LOW): Analysis of scripts/parallel-agent-fanout.py identified an inherent attack surface for Indirect Prompt Injection.
Ingestion points: Untrusted data enters the workflow via url and raw_content fields in the AnalysisState TypedDict.
Boundary markers: No explicit delimiters (e.g., XML tags or specialized tokens) or "ignore embedded instructions" warnings are used when passing content to agent nodes.
Capability inventory: The provided template uses a mock run_agent function with no side effects. However, the architecture is designed to support agents like security_auditor and codebase_analyzer which may have higher-tier capabilities in production environments.
Sanitization: No sanitization or validation logic is present for the ingested external content. Implementers should apply input filtering and boundary markers within individual agent nodes.

langgraph-parallel