## Why Auto Review should remain the effective approval reviewer when settings cross runtime boundaries. A config or app-server round trip must not change the reviewer identity, and delegated work must not silently fall back to user review. This requires both a stable canonical serialized value and propagation of the effective setting. `auto_review` is the canonical value across protocol and app-server output, while `guardian_subagent` remains accepted as backward-compatible input. ## What changed - serialize `ApprovalsReviewer::AutoReview` consistently as `auto_review` across core protocol and app-server v2 - continue accepting `guardian_subagent` when reading existing config or client requests - carry the active turn's approval reviewer into spawned agents - update config/debug expectations and add delegated-task regression coverage ## Scope This does not change Guardian policy or remove compatibility with existing `guardian_subagent` inputs. It preserves the selected reviewer across serialization, config reloads, app-server settings, and delegated task setup. Related Guardian changes are split independently: - #26231 adds denials and soft denials - #26334 retries transient reviewer failures - #26333 reuses narrowly scoped low-risk approvals - #26232 adds TUI denial recovery ## Validation - `just test -p codex-app-server-protocol` (224 passed) - regression coverage for delegated task reviewer propagation - serialization coverage for canonical `auto_review` output and legacy `guardian_subagent` input --------- Co-authored-by: saud-oai <saud@openai.com>
codex-config loader
This module is the canonical place to load and describe Codex configuration layers (user config, CLI/session overrides, cloud-managed config, managed config, and MDM-managed preferences) and to produce:
- An effective merged TOML config.
- Per-key origins metadata (which layer “wins” for a given key).
- Per-layer versions (stable fingerprints) used for optimistic concurrency / conflict detection.
Public surface
Exported from codex_config::loader:
load_config_layers_state(fs, codex_home, cwd_opt, cli_overrides, options, thread_config_loader) -> ConfigLayerStackConfigLayerStackeffective_config() -> toml::Valueorigins() -> HashMap<String, ConfigLayerMetadata>layers_high_to_low() -> Vec<ConfigLayer>with_user_config(user_config) -> ConfigLayerStack
ConfigLayerEntry(one layer’s{name, config, version, disabled_reason};namecarries source metadata)ConfigLoadOptions(user-facing load behavior such as strict config validation)LoaderOverrides(test/override hooks for managed config sources)merge_toml_values(base, overlay)(public helper used elsewhere)
Layering model
Precedence is top overrides bottom:
LegacyManagedConfigTomlFromMdm(MDM-deliveredmanaged_config.toml, while it is being phased out)LegacyManagedConfigTomlFromFile(managed_config.toml, while it is being phased out)SessionFlags(CLI overrides, applied as dotted-path TOML writes)Projectconfig (.codex/config.toml)Userprofile config, when presentUserconfig (config.toml)EnterpriseManagedcloud-managed config bundle layersSystemconfig (/etc/codex/config.tomlor the Windows system config path)
ConfigLayerStack stores layers in the opposite order internally: lowest
precedence first, highest precedence last, so later layers override earlier
layers when folded. Thread config entries supplied by thread_config_loader are
inserted according to their translated ConfigLayerSource precedence.
Layers with a disabled_reason are still surfaced for UI, but are ignored when
computing the effective config and origins metadata. This is what
ConfigLayerStack::effective_config() implements.
Typical usage
Most callers want the effective config plus metadata:
use codex_config::LoaderOverrides;
use codex_config::NoopThreadConfigLoader;
use codex_config::loader::load_config_layers_state;
use codex_exec_server::LOCAL_FS;
use codex_utils_absolute_path::AbsolutePathBuf;
use toml::Value as TomlValue;
let cli_overrides: Vec<(String, TomlValue)> = Vec::new();
let cwd = AbsolutePathBuf::current_dir()?;
let layers = load_config_layers_state(
LOCAL_FS.as_ref(),
&codex_home,
Some(cwd),
&cli_overrides,
LoaderOverrides::default(),
&NoopThreadConfigLoader,
).await?;
let effective = layers.effective_config();
let origins = layers.origins();
let layers_for_ui = layers.layers_high_to_low();
Internal layout
Implementation is split by concern:
state.rs: public types (ConfigLayerEntry,ConfigLayerStack) + merge/origins convenience methods.layer_io.rs: readingconfig.toml, managed config, and managed preferences inputs.overrides.rs: CLI dotted-path overrides → TOML “session flags” layer.merge.rs: recursive TOML merge.fingerprint.rs: stable per-layer hashing and per-key origins traversal.macos.rs: managed preferences integration (macOS only).