The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions facilitate the execution of JavaScript code via the mcp__flow-nexus__sandbox_execute tool, intended for simulating consensus clusters.
[DYNAMIC_EXECUTION]: The skill generates and runs JavaScript logic at runtime to handle consensus rounds, voting mechanisms, and node state management within a sandbox environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from distributed agents, including proposals and voting scores, which are interpolated into coordination logic and solver inputs.
Ingestion points: Data inputs such as proposals, nodeStates, votes, and objectives are defined as primary variables for consensus calculation in SKILL.md.
Boundary markers: No delimiters or instructions to ignore embedded commands are present in the code templates provided for the agent.
Capability inventory: The skill utilizes mcp__flow-nexus__sandbox_execute for script execution and mcp__sublinear-time-solver__solve for algorithmic decision-making.
Sanitization: There is no evidence of input validation or sanitization before external data is passed to the solver or the sandbox environment.

agent-consensus-coordinator