Skip to content

OpenAI's Codex Security CLI and TypeScript SDK for finding, validating, and fixing security vulnerabilities. npm: https://www.npmjs.com/package/@openai/codex-security

fix(plugin): persist large scan usage at completion and failure (#1874)

Michael D'Angeloe2c5a47fac3e
diff --git a/plugins/codex-security/scripts/workbench_validation.py b/plugins/codex-security/scripts/workbench_validation.py
index 7c52f92ed..a03810d02 100644
--- a/plugins/codex-security/scripts/workbench_validation.py
+++ b/plugins/codex-security/scripts/workbench_validation.py
@@ -189,8 +189,6 @@ def _valid_measured_scan_usage(usage: object) -> bool:
def parse_scan_cost(value: str | None) -> str | None:
if value is None:
return None
- if len(value.encode("utf-8")) > 8192:
- raise SystemExit("Scan cost must be no larger than 8 KiB.")
try:
cost = json.loads(value, parse_constant=reject_nonstandard_json_number)
except (TypeError, UnicodeError, ValueError) as exc:
diff --git a/plugins/codex-security/tests/test_workbench_scan_usage.py b/plugins/codex-security/tests/test_workbench_scan_usage.py
index b8f96436d..b81a900a6 100644
--- a/plugins/codex-security/tests/test_workbench_scan_usage.py
+++ b/plugins/codex-security/tests/test_workbench_scan_usage.py
@@ -2152,6 +2152,81 @@ def test_completion_uses_latest_account_snapshot_across_owned_sessions(
)
[email protected]("terminal_status", ["complete", "failed"])
+def test_large_usage_persists_with_terminal_scan_state(
+ tmp_path: Path, terminal_status: str
+) -> None:
+ fixture = _start_scan(tmp_path)
+ usage = {
+ "coverage": "complete",
+ "source": "codex_rollout",
+ "threadCount": 1,
+ **_counts(120, 30, 15),
+ "accountRateLimits": [
+ {
+ "limitId": f"fixture-limit-{index}",
+ "observedAt": fixture.started_at.isoformat(),
+ "primary": {"usedPercent": 6, "windowMinutes": 10080, "resetsAt": 1800000000},
+ "secondary": None,
+ }
+ for index in range(60)
+ ],
+ }
+ cost_json = json.dumps({"usage": usage}, separators=(",", ":"))
+ assert len(cost_json.encode("utf-8")) > 8192
+
+ if terminal_status == "complete":
+ terminal = _complete_scan(fixture, "--cost-json", cost_json)["scan"]
+ else:
+ terminal = fail_scan(
+ fixture.state_dir,
+ fixture.scan_id,
+ "Fixture failure.",
+ "--cost-json",
+ cost_json,
+ environment=fixture.environment,
+ )["scan"]
+
+ assert terminal["progress"]["status"] == terminal_status
+ assert terminal["usage"] == usage
+ stored = scan_command(
+ fixture.state_dir, "get-scan", fixture.scan_id, environment=fixture.environment
+ )["scan"]
+ assert stored["progress"]["status"] == terminal_status
+ assert stored["usage"] == usage
+
+
+def test_completion_collects_large_usage_from_owned_rollouts(tmp_path: Path) -> None:
+ fixture = _start_scan(tmp_path)
+ counted = fixture.started_at + timedelta(microseconds=1)
+ limit_ids = {f"fixture-limit-{index}" for index in range(60)}
+ parent = _rollout(
+ tmp_path,
+ "scan-parent",
+ [
+ _token_event(counted, 120, 15, cached_input_tokens=30),
+ *[_allowance_event(counted, 6, limit_id) for limit_id in sorted(limit_ids)],
+ ],
+ )
+ _state_graph(fixture.environment, {"scan-parent": parent}, [])
+
+ completed = _complete_scan(fixture)["scan"]
+
+ assert completed["progress"]["status"] == "complete"
+ usage = completed["usage"]
+ assert len(json.dumps({"usage": usage}, separators=(",", ":")).encode("utf-8")) > 8192
+ assert usage["coverage"] == "complete"
+ assert usage["threadCount"] == 1
+ assert usage.items() >= _counts(120, 30, 15).items()
+ assert {limit["limitId"] for limit in usage["accountRateLimits"]} == limit_ids
+ assert (
+ scan_command(
+ fixture.state_dir, "get-scan", fixture.scan_id, environment=fixture.environment
+ )["scan"]["usage"]
+ == usage
+ )
+
+
def test_host_final_usage_saves_allowances_before_token_receipts(tmp_path: Path) -> None:
fixture = _start_scan(tmp_path)
arguments = ("--thread-id", "scan-parent", "--turn-id", "turn-scan-parent")