[Security] Conduct threat model and attack surface review

[thealphacubicle]

Overview

Conduct a formal threat model and attack surface review of OpenContext. As the project moves from beta to broader municipal adoption, understanding the attack surface is critical — especially given the MCP protocol's tool-calling model which exposes civic data APIs to AI assistants.

Goals

• Map the full attack surface: Lambda handler, MCP tool endpoints, plugin
  layer, Go client, config.yaml
• Identify injection vectors — can a malicious MCP tool call manipulate
  upstream API queries?
• Review input validation and output sanitization across all three plugins
• Assess risks specific to the fork model — can a misconfigured fork
  expose sensitive data?
• Produce a threat model document with findings and recommended mitigations

Deliverables

• Attack surface map documented
• Input validation reviewed across CKAN, Socrata, ArcGIS plugins
• Injection vector assessment completed
• Fork model risk assessment completed
• Threat model doc committed to docs/ on develop branch
• Follow-on mitigation issues filed for any high/critical findings

Notes

• This is a spike — implementation of mitigations goes in separate issues
• Pay particular attention to SODA3 and ArcGIS REST query parameters
  as injection vectors