The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes potentially untrusted external matrix data (e.g., matrixData, sparseValues, rhsVector) through tools like mcp__sublinear-time-solver__analyzeMatrix. While no explicit malicious instructions are present, the ingestion of large-scale data structures creates a surface for indirect prompt injection if those structures contain embedded natural language instructions meant to influence the agent's output or recommendations.
Ingestion points: matrixData in analyzeMatrix, sparseValues in solve, and rhsVector in estimateEntry.
Boundary markers: None identified in the provided skill definition.
Capability inventory: Includes file/sandbox execution via mcp__flow-nexus__sandbox_execute and network-like coordination via Swarm/Flow Nexus.
Sanitization: No explicit sanitization or validation logic is shown for the incoming data arrays or objects.
[COMMAND_EXECUTION]: The skill facilitates the execution of dynamic Python code within a sandbox environment via mcp__flow-nexus__sandbox_execute. While this is an intended feature for matrix optimization, it allows the agent to run arbitrary logic based on its analysis of potentially untrusted data.

agent-matrix-optimizer