Offline, MCP-native threat modeling: STRIDE + LINDDUN models, DREAD risk, attack trees, and 600+ compliance-framework mappings from a spec or your IaC.
castellan exposes 7 threat-modeling tools via FastMCP with basic descriptions and minimal schema detail. All tools are READ_ONLY and accept simple string or empty inputs. Descriptions are present (avg ~150 chars) but lack actionable guidance on when to use each tool vs. alternatives. Parameter descriptions are minimal (e.g., 'castellan YAML specification' with no format hints). No output schemas are documented, LLMs cannot predict response structure. Error handling returns JSON with 'error' field but provides no recovery guidance. Tool names follow verb_noun pattern (analyze, scan, validate, import, compliance, report, stats) which is good, but lack of composition guidance and missing output documentation significantly limit agent usability.
Build a full STRIDE+LINDDUN threat model from a castellan YAML spec. Returns JSON with threats (risk-scored, with CWE/CAPEC/ATT&CK/OWASP and compliance refs), attack trees, and a summary.
Return the framework control matrix (framework -> category -> control) for a spec.
Convert IaC (Terraform/CFN/K8s/compose) at `path` into a castellan spec (YAML).
Return counts of the bundled threat / requirement / framework catalogs.
Render a full Markdown threat-model report from a castellan spec.
Auto-detect a spec or Infrastructure-as-Code at `path` and model it. Accepts a castellan spec, a Terraform/CloudFormation/Kubernetes/compose file, or a directory. Returns the JSON threat model.
No output schemas documented. LLMs cannot predict response structure (JSON threat model, compliance matrix, markdown report). Agents must guess field names and types, risking extraction errors.
Parameter descriptions lack format/constraint hints. 'castellan YAML specification' does not explain expected structure, required fields, or example format. LLMs cannot validate input before calling.
Error handling returns bare JSON error objects with no recovery guidance. 'SpecError' exceptions are caught and returned as {"error": str(exc)}, but LLMs receive no hint on what to do next (retry, fix input, call different tool).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
Validate a castellan spec; returns {valid, system, elements, flows} or an error.
Tool composition unclear. No guidance on which tool to call first (scan? validate? analyze?), or how outputs chain together. castellan_scan and castellan_analyze both accept specs but differ in input type (path vs. string), distinction not explained.
castellan_library_stats has empty input schema (no parameters), but description does not explain what it returns or when to call it. LLMs cannot determine if this is a discovery tool or a utility.