The Agent Skills Directory

[OBFUSCATION]: The skill utilizes a character substitution technique for documentation and resource URLs, replacing forward slashes with dollar signs (e.g., https:/$flow-nexus.ruv.io$docs). This method is designed to prevent standard link-detection filters from identifying the strings as active URLs.
Evidence: Multiple occurrences throughout SKILL.md, including the Resources, Troubleshooting, and Note sections.
[REMOTE_CODE_EXECUTION]: The skill requires the user to install and run the flow-nexus tool using npx, which involves downloading and executing remote code from the NPM registry at runtime. This tool is a core component provided by the author to manage the E2B sandboxes and training clusters.
Evidence: claude mcp add flow-nexus npx flow-nexus@latest mcp start in the Prerequisites section of SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates a vulnerability surface for indirect prompt injection, as it ingests and processes untrusted data and configurations within its training and prediction workflows.
Ingestion points: External data enters the system through the custom_config parameter in mcp__flow-nexus__neural_deploy_template and the input_data parameter in mcp__flow-nexus__neural_predict_distributed (found in SKILL.md).
Boundary markers: No explicit markers or instructions are provided to the agent to treat these inputs as data only or to ignore any embedded instructions.
Capability inventory: The skill has significant capabilities, including deploying worker nodes, managing training clusters, and publishing models, which could be abused if malicious instructions are successfully injected.
Sanitization: There is no documented validation or sanitization of the input data before it is passed to the underlying Flow Nexus tools.

flow-nexus-neural