The Agent Skills Directory

[NO_CODE]: The skill is entirely composed of Markdown files containing instructions and code examples. It does not include any executable scripts (.py, .js, .sh) or binary files.
[SAFE]: The instructional content emphasizes security and reliability best practices for AI agent development:
[SAFE]: rules/tools-interrupts.md provides patterns for implementing human-in-the-loop approval gates using interrupt() before executing destructive or high-risk tools (e.g., account deletion or large fund transfers).
[SAFE]: rules/routing-retry-loops.md mandates the use of maximum retry counters in agent loops to prevent infinite execution and resource exhaustion.
[SAFE]: rules/state-pydantic.md encourages the use of Pydantic for strict data validation at system boundaries to prevent malformed data from entering the workflow.
[SAFE]: No obfuscation, hardcoded credentials, or malicious network patterns were detected. References to well-known libraries like sentence-transformers for embedding similarity and langgraph for state management are appropriate and used according to standard practices.

langgraph