feat(ai-judge): review sink with telemetry health and fail-closed truth table

- add review.ts sink adapter with session-start review-log toggle detection and a privacy key denylist enforced before delegation
- add judge.ts enforce truth table: allow requires mode, host contract, telemetry health, cohort qualification, owner approval, activation, judgment result, allow verdict, review acknowledgement, and current generation — each independently forces defer with a distinct reason
- route the authorizer callback through the sink and the v0.1 production gate state, which is structurally unreachable and therefore fail-closed
This commit is contained in:
2026-08-17 21:37:48 +08:00
parent 3f3bbb4c28
commit 1d701ca0b0
5 changed files with 405 additions and 9 deletions
@@ -0,0 +1,78 @@
import { describe, expect, it } from "vitest";
import {
evaluateEnforceAuthority,
v01ProductionGateState,
type EnforceGateState,
} from "../src/judge";
const ALL_OPEN: EnforceGateState = {
hostContractPresent: true,
telemetryHealth: "healthy",
cohortQualified: true,
ownerApprovalRecorded: true,
activationRecorded: true,
resultKind: "judgment",
verdict: "allow",
reviewAcknowledged: true,
generationCurrent: true,
mode: "enforce",
};
describe("evaluateEnforceAuthority — every gate independently forces defer", () => {
it("allows only when every gate holds", () => {
expect(evaluateEnforceAuthority(ALL_OPEN)).toEqual({ kind: "allow" });
});
const cases: ReadonlyArray<{
name: string;
patch: Partial<EnforceGateState>;
expectedReason: string;
}> = [
{ name: "shadow mode", patch: { mode: "shadow" }, expectedReason: "mode_shadow" },
{ name: "host contract absent", patch: { hostContractPresent: false }, expectedReason: "host_contract_absent" },
{ name: "telemetry disabled", patch: { telemetryHealth: "disabled" }, expectedReason: "telemetry_disabled" },
{ name: "telemetry write failed", patch: { telemetryHealth: "write_failed" }, expectedReason: "telemetry_write_failed" },
{ name: "telemetry integrity anomaly", patch: { telemetryHealth: "integrity_anomaly" }, expectedReason: "telemetry_integrity_anomaly" },
{ name: "cohort not qualified", patch: { cohortQualified: false }, expectedReason: "cohort_not_qualified" },
{ name: "owner approval absent", patch: { ownerApprovalRecorded: false }, expectedReason: "owner_approval_absent" },
{ name: "activation absent", patch: { activationRecorded: false }, expectedReason: "activation_absent" },
{ name: "preflight result", patch: { resultKind: "preflight_defer" }, expectedReason: "result_preflight_defer" },
{ name: "infrastructure result", patch: { resultKind: "infrastructure_failure" }, expectedReason: "result_infrastructure_failure" },
{ name: "semantic deny verdict", patch: { verdict: "deny" }, expectedReason: "verdict_deny" },
{ name: "semantic defer verdict", patch: { verdict: "defer" }, expectedReason: "verdict_defer" },
{ name: "review unacknowledged", patch: { reviewAcknowledged: false }, expectedReason: "review_unacknowledged" },
{ name: "stale generation", patch: { generationCurrent: false }, expectedReason: "generation_stale" },
];
for (const { name, patch, expectedReason } of cases) {
it(`${name} defers with a distinct reason`, () => {
const state: EnforceGateState = { ...ALL_OPEN, ...patch };
expect(evaluateEnforceAuthority(state)).toEqual({
kind: "defer",
blockedBy: expectedReason,
});
});
}
});
describe("evaluateEnforceAuthority — v0.1 production state", () => {
it("never grants authority for any mode or telemetry state in v0.1", () => {
const modes = ["shadow", "enforce"] as const;
const healths = ["healthy", "disabled", "write_failed", "integrity_anomaly"] as const;
for (const mode of modes) {
for (const health of healths) {
const outcome = evaluateEnforceAuthority(
v01ProductionGateState(mode, health),
);
expect(outcome.kind).toBe("defer");
}
}
});
it("blocks v0.1 enforce on the cohort gate", () => {
const outcome = evaluateEnforceAuthority(
v01ProductionGateState("enforce", "healthy"),
);
expect(outcome).toEqual({ kind: "defer", blockedBy: "cohort_not_qualified" });
});
});
@@ -0,0 +1,79 @@
import { describe, expect, it } from "vitest";
import { createReviewSink, type ReviewSinkDeps } from "../src/review";
import type { AuthorizerLog } from "@gotgenes/pi-permission-system";
function fakeLog(): AuthorizerLog & {
reviews: Array<{ event: string; details: Record<string, unknown> }>;
debugs: Array<{ event: string; details?: Record<string, unknown> }>;
} {
const reviews: Array<{ event: string; details: Record<string, unknown> }> = [];
const debugs: Array<{ event: string; details?: Record<string, unknown> }> = [];
return {
reviews,
debugs,
review: (event, details = {}) => reviews.push({ event, details }),
debug: (event, details) => debugs.push({ event, details }),
};
}
describe("createReviewSink — telemetry health", () => {
it("marks the runtime disabled when the review log toggle is off", () => {
const log = fakeLog();
const deps: ReviewSinkDeps = { log, reviewLogEnabled: false };
const sink = createReviewSink(deps);
expect(sink.health()).toBe("disabled");
});
it("reports healthy when the toggle is on", () => {
const sink = createReviewSink({ log: fakeLog(), reviewLogEnabled: true });
expect(sink.health()).toBe("healthy");
});
});
describe("createReviewSink — privacy denylist at the sink", () => {
it("strips keys matching forbidden patterns before delegation", () => {
const log = fakeLog();
const sink = createReviewSink({ log, reviewLogEnabled: true });
sink.review("ai_bash_judge.result", {
requestId: "req-1",
apiToken: "leak",
sshKey: "leak",
secrets: "leak",
password: "leak",
credentials: "leak",
outputUsage: 10,
});
expect(log.reviews).toEqual([
{
event: "ai_bash_judge.result",
details: { requestId: "req-1", outputUsage: 10 },
},
]);
});
it("passes metadata-only events through unchanged", () => {
const log = fakeLog();
const sink = createReviewSink({ log, reviewLogEnabled: true });
sink.review("ai_bash_judge.result", {
schemaVersion: 1,
requestId: "req-1",
judgeRuntimeId: "abc",
mode: "shadow",
resultKind: "judgment",
});
expect(log.reviews[0]?.details).toEqual({
schemaVersion: 1,
requestId: "req-1",
judgeRuntimeId: "abc",
mode: "shadow",
resultKind: "judgment",
});
});
it("delegates debug writes without stripping", () => {
const log = fakeLog();
const sink = createReviewSink({ log, reviewLogEnabled: true });
sink.debug("ai_bash_judge.exception");
expect(log.debugs).toEqual([{ event: "ai_bash_judge.exception" }]);
});
});