[bokelley]

Unsupported optional controller forcing did not establish the state needed by later observations. Clarify that a stateful force_* call returning UNKNOWN_SCENARIO skips its dependent observations as unavailable coverage, including when forcing is the sole stateful step in its phase. Otherwise a conformant seller can be failed for returning an outcome it was never forced to change.

Register the existing force_scenario_unsupported detail and document its exception to the general sole-stateful-step rule. Preserve independent phases, any_of alternatives, ordinary account exemptions and genuine prerequisite failures. Skipped coverage earns no executed-step or validation credit.

Refs adcontextprotocol/adcp-client#3549. The implementation and regression coverage are in SDK #3561.

Validation

• Protocol changeset scope and Changesets status passed.
• SDK companion: 73 focused tests passed in Docker, including existing cascade coverage and new negative controls; real HTTP MCP force/observe controls passed.
• Independent code/protocol and security reviews passed.
• Normal Docker commit hooks passed: root unit tests, full server unit suite, import/identity guards and both type checks. One unchanged statistical-worker suite failed under parallel load and passed the hook’s automatic isolated retry (34/34).
• CI and the actual triggered storyboard matrix remain merge gates.

Downstream detail: the SDK companion emits its existing capability_prerequisite_unavailable detailed reason, canonical not_applicable, and the originating force step in detail. A protocol-wide downstream token remains a separate dashboard follow-up.

[aao-secretariat]

Ladon verdict: Approve

Approve — clean clarification to the compliance runner-output contract with a correct patch changeset.

What I checked:

• The diff adds a force_scenario_unsupported not_applicable sub-reason and a forced_scenario_cascade rule carving an exception out of sole_stateful_step_exemption, so an unsupported stateful force_* call (UNKNOWN_SCENARIO) skips dependent observations as not_applicable coverage rather than failing a conformant seller.
• Breaking-class hard gate: This is a coverage-gap-not-failure clarification that relaxes grading (a previously-failing conformant seller now grades not_applicable). It does not remove/rename fields, flip optional↔required, or impose a new MUST that a conformant 3.1.0 implementation would fail. It is Normative-non-breaking/clarifying, consistent with the directly-analogous rate-limit-exhaustion precedent (also a patch for a coverage-gap clarification). The breaking-class gate does not fire, so review_decision: REVIEW_REQUIRED does not force escalation here.
• Changeset scope/type: A patch changeset is present and matches precedent; appropriate for a clarification to compliance assets.
• Prose is internally consistent (preserves genuine failures, independent phases, any_of alternatives; dependent skips grade not_applicable, not prerequisite_failed); new detail matches step-disposition.ts/compliance-publication.test.ts; YAML indentation well-formed; dist artifacts untouched.
• gated_paths: false; high_risk: false; no no-auto-approve team match; no Critical/High/Medium findings.

One non-blocking follow-up note: downstream dependent skips are told to "cite the unsupported force step" but no canonical detail token is registered for the downstream case, so detail-filtering dashboards may see free-form prose. Not blocking.

No blocking findings; rows 1–8 do not fire; falls through to row 9 → approve.