Methods and decision trail

ID	Condition	Action	Observable check	Exception	Evidence
M1	Any assessment request	Establish use case, stakeholders, datasets, applications, obligations and authorized scope	Explicit intake and target boundaries	Missing input permits only a provisional plan	S2 §1.2 pp. 8–10; S4 intake; documented workflow with project authorization adaptation
M2	System type is not fully established	Identify all seven deployment features and their yes/no/unknown state	Classification rationale and unresolved questions	A mixed system selects multiple patterns	S1 risk columns H–N; S2 p. 9; S4 multiselect; documented taxonomy with tri-state adaptation
M3	Deployment patterns are known	Union source Yes risks and deduplicate their forward-linked controls	Selected risk IDs, matching columns, control IDs and row citations	Blanks stay unspecified; No does not prove absence	S1 risk columns E and H–N; S4 filtering; documented selection, project provenance implementation
M4	Source and architecture conflict	Preserve original mapping and add a separately labelled architecture review	Source deviation, rationale and owner are explicit	Never silently rewrite source truth	S1 observed 9.1 matrix and asymmetric links; project adaptation
M5	A plausible risk is selected	Define a falsifiable hypothesis, inspect evidence, then test synthetic cases with legitimate comparisons	Expected/observed results, versions, authorization and artifact references	No permission for active tests means evidence review only	S2 pp. 99–103 automated/manual testing; project evidence and authorization adaptation
M6	Tools, MCP, memory or delegation cross trust boundaries	Trace identity, consent, scope, token audience, state ownership and egress at each hop	Boundary-specific denial evidence and attributable effects	Local stdio and remote OAuth have different deployment requirements	S3 component-13 organization; S5 named security sections; documented threats with project checklist adaptation
M7	Evidence supports a control gap	Propose a minimal change, owner, rollback and acceptance test; retest prohibited and legitimate behavior	Evidence of enforcement and stated residual risk	Control description or product presence never counts as a passed test	S1 mitigation descriptions; S2 p. 10 and pp. 100–103; project remediation contract
M8	User asks to understand the failure	Explain mechanism, observed/hypothetical impact, correction and verification in the user’s language	Clear evidence labels, concrete example and source IDs	No verified observation means explain a possible risk, not a confirmed incident	S2 intended audiences pp. 7–8; project communication adaptation

Resolving tensions

Source filtering (M3) offers reproducibility; architecture review (M4) prevents false exclusions. Keep both views with provenance. A guardrail recommendation (S1 DASF-54) does not override the downstream authorization requirements (M6/S5). Traceability (S1 DASF-65) is implemented with observable decisions and effects; private hidden model reasoning is not required. Full coverage never overrides the authorized testing boundary (M1/M5).