Known issue complete

The check is default-on at warning. It runs at verify against every active KnownIssue. A missing evidence kind, a missing introduced_in, or a high-severity KI without owner and mitigation reports a counted warning. It does not block advancement. Opt out with an explicit false. That reports skip, never pass.

Evidence is mandatory. The kind is a choice, ranked. Prefer a bound in-repo test carrying // Reproduces: <KI-slug> (alias // KnownIssue:). Fall back to a resolvable evidence_manifests entry, a reproducer_command, reproduction_steps, or an environment_note only when a bound test is genuinely impossible. The assertion must describe the correct behavior, so the test fails while the bug is present and passes once the fix lands. Encoding today's bug as the expectation certifies the defect.

Origin is also mandatory. The form is a choice: a commit-ish, inception, or unknown. High, critical, or security-relevant KIs must also carry owner, review_date, severity, and a mitigation or remediation. A CVE-shaped surface must also carry a CVSS Base vector that parses. A Go test listed in evidence_manifests without the comment is flagged.

// Reproduces: KI-TRUNC-HEADER
func TestRejectsTruncatedHeader(t *testing.T) {
    _, err := Parse(truncatedHeader)
    if err == nil {
        t.Fatal("parser accepted a truncated header")
    }
}

The comment is the source of truth, the same way // Verifies: binds a test to a requirement. See known issues reviewed for whether the review_date is still current. This hop is whether the record itself is complete enough to replicate. See software problem report for the DEFECT object that is not a KnownIssue.

proof known-issue list
proof known-issue edit KI-TRUNC-HEADER --introduced-in a1b2c3d
proof audit --check known_issue_complete --verbose

Start at rank 1. Write the failing test. Put the comment above it. Record the origin. Do not drop the KI so the hop goes green. A missing reproducer is still a missing reproducer.

The teaching graph is still one KI next to one missing evidence kind. Read the finding. Then pick the first rank you can honestly do.

proof known-issue list --status open
proof known-issue edit KI-TRUNC-HEADER --introduced-in a1b2c3d
proof audit --check known_issue_complete --verbose
proof help known-issue

Close a missing-evidence finding by binding a test that asserts the correct behavior. Record introduced_in. Classify a concrete commit as regression or new_defect. Fill owner and CVSS when the severity demands it. Do not encode the bug as the expected result. Do not treat observed_result: known_issue_reproduced as proof the command ran: the hop checks that the manifest describes a run, not that the run happened. Jama still authors. Proof vs Jama.

Warning severity. Findings are counted warnings, strict-exempt. They do not fail the merge unless you raise the fail level. Zero active KnownIssues is a pass that inspected nothing. An explicit opt-out reports skip, not pass. environment_note is the weakest evidence kind and does not clear the floor alone at high assurance. A complete run record proves the manifest describes an execution, not that the execution happened or that observed_result matches its output. The hop does not rewrite the KI. It does not inspect the implementation. We have not scored this floor against a frozen Jama pack, a Salesforce bulletin, or a Jira project. The loss is named, not scored.

The review-date hop stays on known issues reviewed. The DEFECT object stays on software problem report. The engagement stays on software correctness audit. Jama still authors.

What is known issue complete? Same question. Same URL.

Is this known issues reviewed? No. That hop is whether review_date is still current. This hop is whether the record has evidence and an origin. See known issues reviewed.

Is this a Salesforce known issue? No. Ads known issue is that status page. A hosted vendor bulletin is not this cell.

Is this a software problem report? No. A DEFECT is a different object. See software problem report.

Does a missing-evidence KI fail the merge? No. The check keeps warning severity. A counted warning does not block advancement.

Does an empty known-issues directory fail? No. Zero active KIs is a pass. Put a YAML there when the behavior is confirmed, then bind a test.

Is Proof a Jama alternative for the shall? No. Jama still authors. Proof vs Jama.