Requirements assumptions

The split is old. An assumption is a fact the environment must satisfy before this component can be proved. A guarantee is a behavior this component itself must produce. The words are not severity. A critical safety property is still a guarantee if this code owns it. A low-level input precondition is still an assumption if the caller owns it.

The signed ship decision lives on risk acceptance. A confirmed user-observable miss lives on software problem report as a KnownIssue. The writing rules for the prose live on ASD-STE100. This page is the inventory of req_type=assumption records: owner, review date, or an honest move to KnownIssue, AcceptedRisk, or discharged. It is not a second risk-acceptance lecture.

proof req assumptions list
proof req assumptions status SYS-REQ-014
proof req assumptions review SYS-REQ-014 --status open --external-owner team:ops --review-date 2026-06-30
proof audit --check assumptions_status

Until that review runs, the YAML is a shall wearing assumption clothing. Nobody has said who watches the boundary, or when they look again.

The teaching graph is still a requirement with req_type: assumption. The check still has to see an owner, a review date, a KnownIssue, an AcceptedRisk, or a discharge. The suite can stay. CI needs the file.

proof req assumptions review SYS-REQ-014 --status open \
  --external-owner team:ops \
  --review-date 2026-06-30 \
  --rationale "vendor SLA is outside this repo; ops watches it"
proof req assumptions review SYS-REQ-013 --status known_issue \
  --linked-known-issue KI-1 --external-owner team:protocol
proof req assumptions review SYS-REQ-012 --status discharged \
  --discharged-by evidence/vendor-sla.md

Pick one outcome per flagged assumption. These are alternatives, not steps. An assumption written in risk language has several honest endings, and choosing the wrong one destroys real information. The check is advisory by default: ordinary open assumptions show up in proof audit. It escalates when the owner is missing, the review date is expired, known_issue or accepted_risk has no linked object, or a high-assurance assumption stays open with no discharge path.

Proof does not implement a PMI assumption log, a MathWorks Requirements Table, or SPSS. It does not ping the vendor. A reviewed assumption means a named owner and a date are on the record. It is not a proof that the boundary still holds. It is not a KnownIssue. It is not an AcceptedRisk. Ordinary open assumptions do not fail the build by themselves. Conservative: a missing owner or an expired review is the thing that escalates, not a silent pass. We have not scored this against MathWorks Requirements Table or a PMI register on a frozen pack. The loss is named, not scored.

proof risk accept is a different command. That H1 lives on risk acceptance. Do not treat a clean assumptions list as a signed ship decision.

The STE100 prose H1 stays on ASD-STE100. Jama still authors.

What are requirements assumptions? Same question. Same URL.

What are software assumptions? Same cluster. Same command. Same YAML.

Is this assumption management? Same cluster. PMI still wins the project-planning glossary. Proof stamps owner and date on a requirement file.

Is this risk acceptance? No. That H1 lives on risk acceptance. An assumption is the boundary. AcceptedRisk is the signed ship of a known failure.

Is this a KnownIssue? Only after you move it. If a user can observe the miss, create a KnownIssue. Do not leave a bug dressed as an assumption.

Is Proof MathWorks? No. Jama still authors. Proof takes the authored assumption and asks CI whether an owner and a review date still exist.