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

refactor(workbench): simplify command preparation and transactions (#1390)

mldangelo-oai4c231da1dc0b
diff --git a/plugins/codex-security/scripts/config_preflight.py b/plugins/codex-security/scripts/config_preflight.py
index a0e7127e2..0132f7980 100644
--- a/plugins/codex-security/scripts/config_preflight.py
+++ b/plugins/codex-security/scripts/config_preflight.py
@@ -374,28 +374,6 @@ def config_profile_layer_path(profile: str | None) -> Path | None:
return path if path.is_file() else None
-def parse_multi_agent_v2_enabled(feature_config: Any) -> bool | None:
- if isinstance(feature_config, bool):
- return feature_config
- if not isinstance(feature_config, dict):
- raise ValueError("features.multi_agent_v2 must be a boolean or table")
- if "enabled" not in feature_config:
- return None
- enabled = feature_config["enabled"]
- if not isinstance(enabled, bool):
- raise ValueError("features.multi_agent_v2.enabled must be a boolean")
- return enabled
-
-
-def merge_toml_value(base: Any, overlay: Any) -> Any:
- if isinstance(base, dict) and isinstance(overlay, dict):
- merged = dict(base)
- for key, value in overlay.items():
- merged[key] = merge_toml_value(merged[key], value) if key in merged else value
- return merged
- return overlay
-
-
def lookup_multi_agent_v2_enabled(
*,
config_layers: list[tuple[Path, dict[str, Any]]],
@@ -409,38 +387,29 @@ def lookup_multi_agent_v2_enabled(
return True, enabled, "effective-config"
v2_configs = []
- config_views = list(iter_config_views(config_layers, config_profile))
- for source, config in reversed(config_views):
+ if "features.multi_agent_v2" in effective_config:
+ v2_configs.append(("effective-config", effective_config["features.multi_agent_v2"]))
+ for source, config in iter_config_views(config_layers, config_profile):
found, feature_config = lookup_dotted(config, "features.multi_agent_v2")
if found:
v2_configs.append((source, feature_config))
- if "features.multi_agent_v2" in effective_config:
- v2_configs.append(("effective-config", effective_config["features.multi_agent_v2"]))
-
- if not v2_configs:
- return False, None, None
-
- merged_v2_config: Any = None
- enabled_source: str | None = None
- for index, (source, feature_config) in enumerate(v2_configs):
- tables_merge = isinstance(merged_v2_config, dict) and isinstance(feature_config, dict)
- if index and not tables_merge:
- enabled_source = None
- merged_v2_config = (
- merge_toml_value(merged_v2_config, feature_config) if index else feature_config
- )
- if isinstance(feature_config, bool) or (
- isinstance(feature_config, dict) and "enabled" in feature_config
- ):
- enabled_source = source
-
- enabled = parse_multi_agent_v2_enabled(merged_v2_config)
- return (
- (True, enabled, enabled_source)
- if enabled is not None
- else (True, False, "documented-default")
- )
+ table_override = False
+ for source, feature_config in v2_configs:
+ # Higher tables inherit enabled only through uninterrupted lower tables.
+ if table_override and not isinstance(feature_config, dict):
+ break
+ if isinstance(feature_config, bool):
+ return True, feature_config, source
+ if not isinstance(feature_config, dict):
+ raise ValueError("features.multi_agent_v2 must be a boolean or table")
+ if "enabled" in feature_config:
+ enabled = feature_config["enabled"]
+ if not isinstance(enabled, bool):
+ raise ValueError("features.multi_agent_v2.enabled must be a boolean")
+ return True, enabled, source
+ table_override = True
+ return (True, False, "documented-default") if table_override else (False, None, None)
def resolve_multi_agent_context(
diff --git a/plugins/codex-security/scripts/workbench_db.py b/plugins/codex-security/scripts/workbench_db.py
index 4a28a4e3c..d0bc61497 100644
--- a/plugins/codex-security/scripts/workbench_db.py
+++ b/plugins/codex-security/scripts/workbench_db.py
@@ -42,12 +42,6 @@ import workbench_saved_results as saved_results
import workbench_scan_history as scan_history
import workbench_scan_usage as scan_usage
import workbench_severity as severity
-from filesystem_identity import (
- serialize_filesystem_identity as serialize_filesystem_identity,
-)
-from filesystem_identity import (
- stored_filesystem_identity_matches as stored_filesystem_identity_matches,
-)
from finalize_scan_contract import (
PRODUCER_NAME,
ContractError,
@@ -944,7 +938,7 @@ def _start_prompt_driven_scan(
target_root = scan_target_root(args.scan_root, target)
connection.execute("BEGIN IMMEDIATE")
- try:
+ with connection:
current_target = require_remediation_target(target_path)
current_diff_target = (
require_diff_target(
@@ -1063,10 +1057,6 @@ def _start_prompt_driven_scan(
)
if claimed.rowcount != 1:
raise SystemExit("Codex Security headless scan ownership could not be recorded.")
- connection.commit()
- except BaseException:
- connection.rollback()
- raise
return {**scan_context(connection, scan_id), "startDisposition": "created"}
@@ -1831,24 +1821,6 @@ def coverage_for_comparison(scan: sqlite3.Row) -> dict[str, Any]:
return saved_results.coverage_for_comparison(_WORKBENCH_DB_CONTEXT, scan)
-def preserve_scan_results_locked(connection: sqlite3.Connection, scan_id: str) -> bool:
- return saved_results.preserve_scan_results_locked(_WORKBENCH_DB_CONTEXT, connection, scan_id)
-
-
-def recover_scan_results(
- connection: sqlite3.Connection, args: argparse.Namespace
-) -> dict[str, Any]:
- return saved_results.recover_scan_results(_WORKBENCH_DB_CONTEXT, connection, args)
-
-
-def fail_scan(connection: sqlite3.Connection, args: argparse.Namespace) -> dict[str, Any]:
- return saved_results.fail_scan(_WORKBENCH_DB_CONTEXT, connection, args)
-
-
-def cancel_scan_locked(connection: sqlite3.Connection, args: argparse.Namespace) -> dict[str, Any]:
- return saved_results.cancel_scan_locked(_WORKBENCH_DB_CONTEXT, connection, args)
-
-
def preserve_stopped_results_after_transition(connection: sqlite3.Connection, scan_id: str) -> None:
saved_results.preserve_stopped_results_after_transition(
_WORKBENCH_DB_CONTEXT, connection, scan_id
@@ -1953,7 +1925,7 @@ def request_finding_remediation(
) -> dict[str, Any]:
request_id = require_uuid(args.request_id, "request-id")
action_token = require_uuid(args.action_token, "action-token")
- try:
+ with connection:
occurrence = require_occurrence(connection, args.occurrence_id)
require_finding_open(connection, occurrence["id"])
scan = require_scan(connection, occurrence["scan_id"])
@@ -2042,10 +2014,6 @@ def request_finding_remediation(
timestamp,
),
)
- connection.commit()
- except BaseException:
- connection.rollback()
- raise
return scan_context(connection, occurrence["scan_id"])
@@ -2054,7 +2022,7 @@ def request_finding_remediation_action(
) -> dict[str, Any]:
request_id = require_uuid(args.request_id, "request-id")
action_token = require_uuid(args.action_token, "action-token")
- try:
+ with connection:
occurrence = require_occurrence(connection, args.occurrence_id)
require_finding_open(connection, occurrence["id"])
scan = require_scan(connection, occurrence["scan_id"])
@@ -2118,10 +2086,6 @@ def request_finding_remediation_action(
raise SystemExit(
"This remediation request changed. Refresh it before recording an update."
)
- connection.commit()
- except BaseException:
- connection.rollback()
- raise
return scan_context(connection, occurrence["scan_id"])
@@ -2131,7 +2095,7 @@ def claim_finding_remediation_resend(
request_id = require_uuid(args.request_id, "request-id")
action_token = require_uuid(args.action_token, "action-token")
connection.execute("BEGIN IMMEDIATE")
- try:
+ with connection:
timestamp = now()
occurrence = require_occurrence(connection, args.occurrence_id)
require_finding_open(connection, occurrence["id"])
@@ -2198,10 +2162,6 @@ def claim_finding_remediation_resend(
unavailable = "This remediation host request is still owned by another panel. Retry after its lease expires."
if updated.rowcount != 1:
raise SystemExit(unavailable)
- connection.commit()
- except BaseException:
- connection.rollback()
- raise
result = scan_context(connection, occurrence["scan_id"])
result["actionToken"] = claimed_token
return result
@@ -2377,10 +2337,6 @@ def set_finding_remediation(
return scan_context(connection, occurrence["scan_id"])
-def export_findings(connection: sqlite3.Connection, args: argparse.Namespace) -> dict[str, Any]:
- return publication.export_findings(_WORKBENCH_PUBLICATION_CONTEXT, connection, args)
-
-
def require_reviewed_patch_applied(
scan: sqlite3.Row, remediation: sqlite3.Row, patch_path: str
) -> str | None:
@@ -3282,10 +3238,7 @@ def read_json_object(path: Path) -> dict[str, Any]:
return payload
-_WORKBENCH_PUBLICATION_CONTEXT = SimpleNamespace(**globals())
-
-
-_WORKBENCH_DB_CONTEXT = SimpleNamespace(**globals())
+_WORKBENCH_PUBLICATION_CONTEXT = _WORKBENCH_DB_CONTEXT = SimpleNamespace(**globals())
def main() -> None:
@@ -3436,11 +3389,11 @@ def main() -> None:
elif args.command == "cancel-scan":
result = saved_results.cancel_scan(_WORKBENCH_DB_CONTEXT, connection, args)
elif args.command == "fail-scan":
- result = fail_scan(connection, args)
+ result = saved_results.fail_scan(_WORKBENCH_DB_CONTEXT, connection, args)
elif args.command == "preserve-scan-results":
result = saved_results.preserve_scan_results(_WORKBENCH_DB_CONTEXT, connection, args)
elif args.command == "recover-scan-results":
- result = recover_scan_results(connection, args)
+ result = saved_results.recover_scan_results(_WORKBENCH_DB_CONTEXT, connection, args)
elif args.command == "write-scan-draft":
result = saved_results.write_scan_draft(_WORKBENCH_DB_CONTEXT, connection, args)
elif args.command == "save-scan-artifact":
@@ -3481,7 +3434,7 @@ def main() -> None:
_WORKBENCH_PUBLICATION_CONTEXT, connection, args
)
elif args.command == "export-findings":
- result = export_findings(connection, args)
+ result = publication.export_findings(_WORKBENCH_PUBLICATION_CONTEXT, connection, args)
elif args.command == "database-info":
result = {"databasePath": str(database_path())}
elif args.command == "severity-classification":
diff --git a/plugins/codex-security/scripts/workbench_scan_start.py b/plugins/codex-security/scripts/workbench_scan_start.py
index cac00420f..e4aa20191 100644
--- a/plugins/codex-security/scripts/workbench_scan_start.py
+++ b/plugins/codex-security/scripts/workbench_scan_start.py
@@ -5,6 +5,7 @@ from __future__ import annotations
import argparse
import json
import os
+import re
import sqlite3
import sys
import tempfile
@@ -26,10 +27,7 @@ from workbench_validation import optional_text
def safe_segment(value: str) -> str:
- segment = "".join(
- character if character.isalnum() or character in "._-" else "-" for character in value
- )
- return segment.strip("-") or "scan"
+ return re.sub(r"[^\w.-]", "-", value).strip("-") or "scan"
def compact_timestamp() -> str:
diff --git a/plugins/codex-security/scripts/workbench_target.py b/plugins/codex-security/scripts/workbench_target.py
index 9918cc84b..04fed9335 100644
--- a/plugins/codex-security/scripts/workbench_target.py
+++ b/plugins/codex-security/scripts/workbench_target.py
@@ -605,12 +605,9 @@ def source_directory_snapshot_paths(target: Path) -> list[Path]:
def directory_content_digest(
target: Path, *, excluded: tuple[Path, ...] = (), include_ignored: bool = False
) -> str:
- excluded_relative = []
- for path in excluded:
- try:
- excluded_relative.append(path.relative_to(target))
- except ValueError:
- continue
+ excluded_relative = [
+ path.relative_to(target) for path in excluded if path.is_relative_to(target)
+ ]
paths = (
source_directory_snapshot_paths(target)
if include_ignored
@@ -622,10 +619,7 @@ def directory_content_digest(
update_digest_field(digest, b"format", b"codex-security-directory/v1")
for path in paths:
relative_path = path.relative_to(target)
- if any(
- relative_path == excluded_path or excluded_path in relative_path.parents
- for excluded_path in excluded_relative
- ):
+ if any(relative_path.is_relative_to(path) for path in excluded_relative):
continue
try:
metadata = path.lstat()
@@ -675,12 +669,9 @@ def directory_snapshot_regular_file_count(target: Path) -> int:
def copy_directory_excluding(source: Path, destination: Path, excluded: tuple[Path, ...]) -> None:
- excluded_relative = []
- for path in excluded:
- try:
- excluded_relative.append(path.relative_to(source))
- except ValueError:
- continue
+ excluded_relative = [
+ path.relative_to(source) for path in excluded if path.is_relative_to(source)
+ ]
def ignored(directory: str, names: list[str]) -> list[str]:
relative = Path(directory).relative_to(source)
@@ -707,19 +698,13 @@ def copy_git_worktree_files(source: Path, destination: Path, excluded: tuple[Pat
)
if listed is None:
raise SystemExit("Could not inspect files in the selected Git working tree.")
- excluded_relative = []
- for path in excluded:
- try:
- excluded_relative.append(path.relative_to(repository))
- except ValueError:
- continue
+ excluded_relative = [
+ path.relative_to(repository) for path in excluded if path.is_relative_to(repository)
+ ]
destination.mkdir()
for raw_path in sorted(path for path in listed.split(b"\0") if path):
relative = Path(os.fsdecode(raw_path))
- if any(
- relative == excluded_path or excluded_path in relative.parents
- for excluded_path in excluded_relative
- ):
+ if any(relative.is_relative_to(path) for path in excluded_relative):
continue
source_path = repository / relative
try:
diff --git a/plugins/codex-security/tests/test_deep_scan_successful_publication.py b/plugins/codex-security/tests/test_deep_scan_successful_publication.py
index 9dcadb55e..4fbb5376c 100644
--- a/plugins/codex-security/tests/test_deep_scan_successful_publication.py
+++ b/plugins/codex-security/tests/test_deep_scan_successful_publication.py
@@ -339,7 +339,8 @@ def test_stopped_deep_scan_still_salvages_saved_findings(
result.write_text("{interrupted worker output")
result_bytes = result.read_bytes()
- stopped = workbench_api["fail_scan"](
+ stopped = workbench_api["saved_results"].fail_scan(
+ workbench_api["_WORKBENCH_DB_CONTEXT"],
workbench_db,
Namespace(
scan_id=scan.scan_id, claim_token=None, cost_json=None, message="Scan interrupted."
@@ -359,8 +360,8 @@ def test_stopped_deep_scan_still_salvages_saved_findings(
artifact_names = ("scan-manifest.json", "findings.json", "coverage.json")
published = {name: (scan.scan_dir / name).read_bytes() for name in artifact_names}
- recovered = workbench_api["recover_scan_results"](
- workbench_db, Namespace(scan_id=scan.scan_id)
+ recovered = workbench_api["saved_results"].recover_scan_results(
+ workbench_api["_WORKBENCH_DB_CONTEXT"], workbench_db, Namespace(scan_id=scan.scan_id)
)["scan"]
assert recovered["findingCount"] == len(expected_summaries)
assert {name: (scan.scan_dir / name).read_bytes() for name in artifact_names} == published
@@ -402,7 +403,8 @@ def test_stopped_deep_scan_ignores_non_reducer_sources_without_coverage(
source_bytes = source_path.read_bytes()
source_relative = source_path.relative_to(scan.scan_dir).as_posix()
- stopped = workbench_api["fail_scan"](
+ stopped = workbench_api["saved_results"].fail_scan(
+ workbench_api["_WORKBENCH_DB_CONTEXT"],
workbench_db,
Namespace(
scan_id=scan.scan_id, claim_token=None, cost_json=None, message="Scan interrupted."
@@ -424,8 +426,8 @@ def test_stopped_deep_scan_ignores_non_reducer_sources_without_coverage(
artifact_names = ("scan-manifest.json", "findings.json", "coverage.json")
published = {name: (scan.scan_dir / name).read_bytes() for name in artifact_names}
- recovered = workbench_api["recover_scan_results"](
- workbench_db, Namespace(scan_id=scan.scan_id)
+ recovered = workbench_api["saved_results"].recover_scan_results(
+ workbench_api["_WORKBENCH_DB_CONTEXT"], workbench_db, Namespace(scan_id=scan.scan_id)
)["scan"]
assert recovered["findingCount"] == 1
diff --git a/plugins/codex-security/tests/test_windows_report_e2e.py b/plugins/codex-security/tests/test_windows_report_e2e.py
index 6a578a48d..84404e84b 100644
--- a/plugins/codex-security/tests/test_windows_report_e2e.py
+++ b/plugins/codex-security/tests/test_windows_report_e2e.py
@@ -49,8 +49,10 @@ def test_workbench_completion_and_exports_use_windows_file_backend(tmp_path: Pat
connection,
SimpleNamespace(claim_token=None, cost_json=None, scan_id=scan_id),
)["scan"]
- exported = namespace["export_findings"](
- connection, SimpleNamespace(scan_id=scan_id, format="csv")
+ exported = namespace["publication"].export_findings(
+ namespace["_WORKBENCH_PUBLICATION_CONTEXT"],
+ connection,
+ SimpleNamespace(scan_id=scan_id, format="csv"),
)
assert completed["progress"]["status"] == "complete"
diff --git a/plugins/codex-security/tests/test_workbench_db.py b/plugins/codex-security/tests/test_workbench_db.py
index 7ffb4fa89..ce1bf87e3 100644
--- a/plugins/codex-security/tests/test_workbench_db.py
+++ b/plugins/codex-security/tests/test_workbench_db.py
@@ -1312,7 +1312,7 @@ def test_completed_finding_triage_and_remediation_persist(
def test_filesystem_identity_serialization_supports_windows_stat_values() -> None:
- namespace = runpy.run_path(str(SCRIPT), run_name="codex_security_workbench_db")
+ namespace = runpy.run_path(str(SCRIPT.with_name("filesystem_identity.py")))
serialize_identity = namespace["serialize_filesystem_identity"]
identity_matches = namespace["stored_filesystem_identity_matches"]
windows_device_id = (1 << 64) - 1
diff --git a/plugins/codex-security/tests/test_workbench_prompt_only_scan.py b/plugins/codex-security/tests/test_workbench_prompt_only_scan.py
index a29d1ab24..2ec1e131a 100644
--- a/plugins/codex-security/tests/test_workbench_prompt_only_scan.py
+++ b/plugins/codex-security/tests/test_workbench_prompt_only_scan.py
@@ -263,7 +263,8 @@ def test_setup_scan_reuses_checked_target_metadata(tmp_path: Path) -> None:
"SELECT target_device, target_inode FROM scans WHERE id = ?",
(scan_id,),
).fetchone()
- serialize_identity = start_globals["serialize_filesystem_identity"]
+ identity_helpers = runpy.run_path(str(SCRIPT.with_name("filesystem_identity.py")))
+ serialize_identity = identity_helpers["serialize_filesystem_identity"]
assert identity == (
serialize_identity(metadata.st_dev),
serialize_identity(metadata.st_ino),
diff --git a/sdk/typescript/tests-ts/stopped-scan-results.test.ts b/sdk/typescript/tests-ts/stopped-scan-results.test.ts
index 290b28996..e0b562e68 100644
--- a/sdk/typescript/tests-ts/stopped-scan-results.test.ts
+++ b/sdk/typescript/tests-ts/stopped-scan-results.test.ts
@@ -80,7 +80,7 @@ const stoppedScanProbe = [
" connection = workbench_db.connect()",
" original_write = workbench_db.saved_results._write_prepared_scan_finalization",
" workbench_db.saved_results._write_prepared_scan_finalization = lambda prepared, *, projection_warnings=None: (_ for _ in ()).throw(OSError('synthetic cancellation publication failure'))",
- " workbench_db.cancel_scan_locked(connection, argparse.Namespace(scan_id=scan_id, thread_id=None))",
+ " workbench_db.saved_results.cancel_scan_locked(workbench_db._WORKBENCH_DB_CONTEXT, connection, argparse.Namespace(scan_id=scan_id, thread_id=None))",
" workbench_db.saved_results._write_prepared_scan_finalization = original_write",
" connection.close()",
" run('preserve-scan-results', '--scan-id', scan_id, '--thread-id', 'stopped-result-owner')",
@@ -109,12 +109,12 @@ const stoppedScanProbe = [
" workbench_db.saved_results._write_prepared_scan_finalization = lambda prepared, *, projection_warnings=None: (_ for _ in ()).throw(OSError('synthetic publication failure'))",
" first_failed = False",
" try:",
- " workbench_db.preserve_scan_results_locked(connection, scan_id)",
+ " workbench_db.saved_results.preserve_scan_results_locked(workbench_db._WORKBENCH_DB_CONTEXT, connection, scan_id)",
" except OSError:",
" first_failed = True",
" frozen_after_failure = connection.execute('SELECT retained_source_digests_json FROM scans WHERE id = ?', (scan_id,)).fetchone()[0]",
" workbench_db.saved_results._write_prepared_scan_finalization = original_write",
- " retry_published = workbench_db.preserve_scan_results_locked(connection, scan_id)",
+ " retry_published = workbench_db.saved_results.preserve_scan_results_locked(workbench_db._WORKBENCH_DB_CONTEXT, connection, scan_id)",
" frozen_after_success = connection.execute('SELECT retained_source_digests_json FROM scans WHERE id = ?', (scan_id,)).fetchone()[0]",
" final_manifest = json.loads(manifest_path.read_text(encoding='utf-8'))",
" final_findings = json.loads((scan_dir / 'findings.json').read_text(encoding='utf-8'))['findings']",