feat: gate Code Agent session registry

Implements #317 session-gated read-only runner phase with structured long-lived Codex stdio blockers.
This commit is contained in:
Lyon
2026-05-23 18:19:58 +08:00
committed by GitHub
parent 97601c9b9d
commit d7e84f359b
13 changed files with 1052 additions and 96 deletions
+82 -7
View File
@@ -1,5 +1,6 @@
const trustedLiveProviders = new Set(["openai-responses", "codex-cli", "codex-readonly-runner"]);
const trustedRunnerProviders = new Set(["codex-readonly-runner"]);
const readonlySessionCapabilityLevels = new Set(["read-only-tools", "read-only-session-tools"]);
const untrustedProviderPattern = /\b(?:echo|mock|fixture|stub|sample)\b/iu;
const blockedCodeAgentUiLabels = new Set(["发送失败", "服务受阻", "BLOCKED 凭证缺口"]);
const timeoutPattern = /\b(?:timeout|timed out|abort|aborted|超时|请求超过)\b/iu;
@@ -86,6 +87,10 @@ export function classifyCodexRunnerCapability(payload, { httpStatus = null } = {
? payload.runner.runnerLimitations.filter((item) => typeof item === "string")
: [];
const sessionReuse = payload?.sessionReuse && typeof payload.sessionReuse === "object" ? payload.sessionReuse : null;
const session = payload?.session && typeof payload.session === "object" ? payload.session : null;
const longLivedSessionGate = payload?.longLivedSessionGate && typeof payload.longLivedSessionGate === "object"
? payload.longLivedSessionGate
: null;
if (provider === "openai-responses" || runnerKind === "openai-responses-fallback") {
return {
@@ -101,11 +106,15 @@ export function classifyCodexRunnerCapability(payload, { httpStatus = null } = {
if (payload?.status !== "completed") missing.push("completed status");
if (!trustedRunnerProviders.has(provider)) missing.push("runner provider");
if (runnerKind !== "hwlab-readonly-runner") missing.push("runner.kind");
if (capabilityLevel !== "read-only-tools") missing.push("capabilityLevel=read-only-tools");
if (!readonlySessionCapabilityLevels.has(capabilityLevel)) missing.push("capabilityLevel=read-only-session-tools");
if (!workspace) missing.push("workspace");
if (sandbox !== "read-only") missing.push("sandbox=read-only");
if (sessionMode !== "controlled-readonly-session-registry") missing.push("sessionMode=controlled-readonly-session-registry");
if (implementationType !== "controlled-readonly-session-registry") missing.push("implementationType=controlled-readonly-session-registry");
if (!session) missing.push("session");
if (session && !["idle", "ready", "busy"].includes(session.status)) missing.push("session.status active/idle");
if (session && typeof session.idleTimeoutMs !== "number") missing.push("session.idleTimeoutMs");
if (session && !session.lastTraceId) missing.push("session.lastTraceId");
if (!sessionReuse) missing.push("sessionReuse");
if (toolCalls.length === 0) missing.push("toolCalls");
if (!runnerTrace) missing.push("runnerTrace");
@@ -114,6 +123,7 @@ export function classifyCodexRunnerCapability(payload, { httpStatus = null } = {
if (payload?.runner?.codexStdio !== false) missing.push("runner.codexStdio=false");
if (payload?.runner?.writeCapable !== false) missing.push("runner.writeCapable=false");
if (payload?.runner?.durableSession !== false) missing.push("runner.durableSession=false");
if (!longLivedSessionGate) missing.push("longLivedSessionGate");
if (missing.length > 0) {
return {
@@ -125,6 +135,16 @@ export function classifyCodexRunnerCapability(payload, { httpStatus = null } = {
};
}
if (longLivedSessionGate?.status === "pass") {
return {
status: "blocked",
level: "BLOCKED/runner-gate",
blocker: "unexpected-long-lived-green",
capabilityPass: false,
reason: "read-only session registry responses must not claim the full long-lived Codex stdio gate."
};
}
if (toolCalls.some((tool) => tool.status !== "completed")) {
return {
status: "blocked",
@@ -137,10 +157,11 @@ export function classifyCodexRunnerCapability(payload, { httpStatus = null } = {
return {
status: "pass",
level: "#275 read-only Codex runner capability pass",
level: "#317 read-only session registry partial pass",
blocker: null,
capabilityPass: true,
reason: "completed read-only runner response includes workspace, sandbox, toolCalls, skills, and runnerTrace evidence"
longLivedSession: false,
reason: "completed read-only session-registry response includes session, workspace, sandbox, toolCalls, skills, runnerTrace, and a blocked long-lived Codex stdio gate"
};
}
@@ -240,9 +261,11 @@ export function summarizeCodeAgentPayload(payload, { httpStatus = null, traceId
runnerTrace: null,
sessionMode: null,
sessionReuse: null,
session: null,
implementationType: null,
runnerLimitations: [],
codexStdioFeasibility: null
codexStdioFeasibility: null,
longLivedSessionGate: null
};
}
@@ -285,9 +308,11 @@ export function summarizeCodeAgentPayload(payload, { httpStatus = null, traceId
runnerTrace: summarizeRunnerTrace(payload.runnerTrace),
sessionMode: stringOrNull(payload.sessionMode),
sessionReuse: summarizeSessionReuse(payload.sessionReuse),
session: summarizeSession(payload.session),
implementationType: stringOrNull(payload.implementationType),
runnerLimitations: summarizeStringList(payload.runnerLimitations),
codexStdioFeasibility: summarizeCodexStdioFeasibility(payload.codexStdioFeasibility)
codexStdioFeasibility: summarizeCodexStdioFeasibility(payload.codexStdioFeasibility),
longLivedSessionGate: summarizeLongLivedSessionGate(payload.longLivedSessionGate)
};
}
@@ -310,9 +335,11 @@ export function summarizeCodeAgentPayload(payload, { httpStatus = null, traceId
runnerTrace: summarizeRunnerTrace(payload.runnerTrace),
sessionMode: stringOrNull(payload.sessionMode),
sessionReuse: summarizeSessionReuse(payload.sessionReuse),
session: summarizeSession(payload.session),
implementationType: stringOrNull(payload.implementationType),
runnerLimitations: summarizeStringList(payload.runnerLimitations),
codexStdioFeasibility: summarizeCodexStdioFeasibility(payload.codexStdioFeasibility)
codexStdioFeasibility: summarizeCodexStdioFeasibility(payload.codexStdioFeasibility),
longLivedSessionGate: summarizeLongLivedSessionGate(payload.longLivedSessionGate)
};
}
@@ -409,6 +436,9 @@ export function inspectRealCompletionEvidence(payload) {
if (provider === "openai-responses" && payload?.capabilityLevel && payload.capabilityLevel !== "text-chat-only") {
return { ok: false, reason: "openai-responses cannot claim runner capabilityLevel" };
}
if (provider === "codex-readonly-runner" && payload?.longLivedSessionGate?.status === "pass") {
return { ok: false, reason: "read-only runner cannot claim long-lived session gate pass" };
}
if (untrustedProviderPattern.test(`${provider} ${backend}`)) {
return { ok: false, reason: `provider/backend looks non-real: ${payload.provider}/${payload.backend}` };
}
@@ -521,7 +551,32 @@ function summarizeSessionReuse(value) {
reused: value.reused === true,
turn: typeof value.turn === "number" ? value.turn : null,
previousTurns: typeof value.previousTurns === "number" ? value.previousTurns : null,
workspace: stringOrNull(value.workspace)
workspace: stringOrNull(value.workspace),
status: stringOrNull(value.status),
idleTimeoutMs: typeof value.idleTimeoutMs === "number" ? value.idleTimeoutMs : null,
lastTraceId: stringOrNull(value.lastTraceId)
};
}
function summarizeSession(value) {
if (!value || typeof value !== "object") return null;
return {
sessionId: stringOrNull(value.sessionId),
conversationId: stringOrNull(value.conversationId),
status: stringOrNull(value.status),
workspace: stringOrNull(value.workspace),
sandbox: stringOrNull(value.sandbox),
runnerKind: stringOrNull(value.runnerKind),
capabilityLevel: stringOrNull(value.capabilityLevel),
implementationType: stringOrNull(value.implementationType),
createdAt: stringOrNull(value.createdAt),
updatedAt: stringOrNull(value.updatedAt),
idleTimeoutMs: typeof value.idleTimeoutMs === "number" ? value.idleTimeoutMs : null,
lastTraceId: stringOrNull(value.lastTraceId),
turn: typeof value.turn === "number" ? value.turn : null,
longLivedSession: value.longLivedSession === true,
codexStdio: value.codexStdio === true,
secretMaterialStored: value.secretMaterialStored === true
};
}
@@ -542,6 +597,26 @@ function summarizeCodexStdioFeasibility(value) {
};
}
function summarizeLongLivedSessionGate(value) {
if (!value || typeof value !== "object") return null;
return {
status: stringOrNull(value.status),
pass: value.pass === true,
requiredCapability: stringOrNull(value.requiredCapability),
currentCapability: stringOrNull(value.currentCapability),
sessionId: stringOrNull(value.sessionId),
sessionStatus: stringOrNull(value.sessionStatus),
runnerKind: stringOrNull(value.runnerKind),
provider: stringOrNull(value.provider),
blockers: Array.isArray(value.blockers)
? value.blockers.map((blocker) => ({
code: stringOrNull(blocker.code),
sourceIssue: stringOrNull(blocker.sourceIssue)
}))
: []
};
}
function summarizeStringList(value) {
return Array.isArray(value) ? value.filter((item) => typeof item === "string") : [];
}