## Why
Signaling an MCP server's process group can fail with `EPERM` on macOS,
preventing owned descendants from being terminated.
## What changed
- When a group signal is denied, enumerate the group's processes and signal
members that still belong to the expected process group.
- Apply the fallback to both `SIGTERM` and `SIGKILL` during MCP stdio server
cleanup, while retaining the existing group signaling behavior elsewhere.
- Reject invalid process group IDs and avoid signaling processes that have
moved to another group.
## Testing
Added macOS tests covering cleanup with a live or exited group leader,
`SIGTERM`-resistant descendants, escalation to `SIGKILL`, and unsafe group IDs.
GitOrigin-RevId: db722475b94dece74244ce8109b3e92f6cfb202a
## Why
The 4 KiB aggregate limit can truncate source metadata before it reaches the
model through the tool search description.
## What changed
Raise the aggregate source-description limit to 512 KiB while retaining the
existing character-boundary-safe truncation and bounded output.
## Testing
Expand unit and RMCP round-trip coverage to verify that substantially longer
Unicode descriptions remain complete within the new limit.
GitOrigin-RevId: 002760cdd2e8741e531b45f75f8624ef84334d24
## What changed
- Propagate the optional `modelSpecialty` field through the model catalog and
app-server `model/list` response.
- When a newly selected model has the `cyber` specialty, default an active TUI
thread to workspace-write permissions with on-request approval. Use auto
review when available, otherwise keep the user as reviewer, while respecting
configured permission requirements.
- Show a notice when auto review is applied and strengthen the full-access
warning for cyber models.
- Preserve explicitly selected permissions when only reasoning settings change.
## Testing
- Cover auto-review defaults, requirement-aware fallback behavior, permission
preservation across reasoning changes, and the new warning notice.
GitOrigin-RevId: 8f25b0082d3051d9ff91f7355921236d0f82f758
## What changed
- Treat different non-empty descriptions for the same tool namespace as a
collision when `error_on_tool_collisions` is enabled, including deferred
tools.
- Continue allowing multiple tools to share a namespace when their descriptions
match or are empty.
- Preserve the first non-empty namespace description when strict collision
checking is disabled.
## Testing
- Cover strict collisions across direct and deferred tools, valid shared
namespaces, relaxed description merging, and failure before model sampling.
GitOrigin-RevId: 788a2ae0eccb64c97995af7a208d9abb79edf2d0
## What changed
- Build filesystem and permission context from the primary turn environment's permission profile, working directory, and workspace roots, falling back to the thread context when no primary environment is available.
- Build capability-discovery sandbox contexts from each environment's permission profile and treat discovery as restricted when any selected environment has restricted filesystem access.
## Testing
- Add regression tests covering environment-specific capability discovery and permission context updates.
GitOrigin-RevId: 6bbbd9b88aafa5e6583971ce4e427fc7557f2d93
## What changed
- Derive sandboxing, approval, and permission-profile environment variables from the selected turn environment for shell, unified exec, `apply_patch`, image viewing, and network approval flows.
- Materialize project-root permissions against every workspace root in that environment, allowing `apply_patch` to write to secondary workspace roots while preserving protected paths.
- Carry the selected environment's permission profile into spawned or resumed agents and Guardian permission context.
## Testing
- Add an integration test covering `apply_patch` writes in a secondary workspace root.
- Extend tool, Guardian, and multi-agent tests to distinguish environment permissions from thread-level permissions.
GitOrigin-RevId: e22da8974bcd3de1e724978b1a7790307a51e4bc
## What changed
- Expect tool collisions to report names canonicalized into the default
`functions` namespace.
- Remove the obsolete test that allowed an explicitly namespaced
`functions.shell_command` external tool.
GitOrigin-RevId: 9a517f49c027ffe5b811f6668dca91246f4925a6
## What changed
- Include the current permission profile in each turn environment's configuration.
- Refresh retained environment configurations when session settings change, while preserving the configuration already captured by active turns.
- Keep environment connection and shell snapshot resolution reusable across configuration updates.
## Testing
- Cover permission profile updates applied through both session settings updates and per-turn updates.
- Verify that pending environment resolutions use the configuration captured by their turn and that configuration-only updates do not retry failed environments.
GitOrigin-RevId: 923694f0729b87a118d7635ade3a8ca9bcd4e7b1
## What changed
- Track Agent Plugin manifests through plugin, skill, and MCP loading so their capabilities use format-specific behavior without changing legacy plugins.
- Discover only direct-child skills, exclude app and hook capabilities, isolate MCP data, and reject MCP configuration files that are non-regular or resolve outside the plugin root.
- Bound model-visible skill instructions, plugin instructions, MCP descriptions, schemas, individual tools, and the aggregate Agent Plugin MCP tool set.
- Stop MCP and OAuth redirects when Agent Plugins send configured or authorization headers, while retaining existing redirect behavior for legacy MCP servers.
## Testing
- Add coverage for capability filtering, skill discovery, isolated MCP data and reserved-path expansion, unsafe MCP configuration files, context limits, and redirect handling.
GitOrigin-RevId: c9af66b051269f3226628ca280a58d32c808c38f
## What changed
- Group top-level function and custom tool definitions into a single `functions` namespace for Responses Lite providers that support namespaced tools.
- Normalize missing, empty, and explicit `functions` namespaces to the same tool identity across registration, routing, lifecycle hooks, configuration, and tool search.
- Keep default tool names unprefixed in code mode, display output, and dispatch traces while preserving explicit non-default namespaces.
## Testing
- Add coverage for Responses Lite serialization, tool search results, namespace normalization and collision handling, routing, lifecycle events, and code-mode namespace policies.
GitOrigin-RevId: d48414005b5d22d39b11a198e19c47814d3a19f2
## What changed
- Track the first duplicate effective tool name while assembling the tool registry, including collisions introduced by external tools, code mode, and tool search.
- When `[features.tool_registry].error_on_tool_collisions` is enabled, fail the turn with a `duplicate tool: <namespace>.<name>` error before sending a model request.
- Continue allowing identical tool names in different namespaces, and preserve the existing collision behavior when strict checking is disabled.
## Testing
- Cover registry collision tracking, strict tool planning across tool sources, namespace isolation, failure before sampling, and pre-sampling compaction.
GitOrigin-RevId: 4cf83538fb513cec0b9c8b567780caaaadd3243d
## What changed
- Key cached skill snapshots by filesystem and plugin snapshot identity so compatible config and working-directory loads can share results without reusing stale plugin data.
- Coalesce concurrent loads for the same cache key and replace the cached entry on forced reload.
- Clear both plugin and skill caches when `skills/list` forces a reload, and bypass working-directory caching when effective plugin roots are present.
## Testing
- Cover concurrent cache sharing, filesystem isolation, forced reloads, and refreshed plugin skill metadata in host service and app server tests.
GitOrigin-RevId: 03fed3b40d45bb29206d5a3c3e78f06df04dbb1e
## What changed
- Include top-level freeform tools in the tool-search index and mark them for deferred loading.
- Serialize searched freeform tools as Responses API `custom` tools and convert them back into executable tool specs after discovery.
- Cover indexing and the end-to-end search, load, and custom-call routing flow.
GitOrigin-RevId: 124a7705876067542e1bd609546adc9a8d0ef67e
## Why
Clients using `thread/read` with `includeTurns: true` need the legacy full-history
view even when a thread stores its history in paginated form.
## What changed
- Reconstruct full projected turns from paginated history for both stored and
loaded threads.
- Keep `thread/turns/list` and `thread/items/list` as the preferred incremental
history APIs.
## Testing
- Cover `thread/read(includeTurns=true)` before and after a paginated thread is
loaded.
GitOrigin-RevId: 6efdc04e5ed3cad2ff2078d8103933082c7921db
## What changed
- Add the public asynchronous `ModelsCache` contract and cache entry/error types.
- Let model providers and `OpenAiModelsManager` accept caller-provided cache implementations while retaining the existing file-backed cache by default.
- Treat cache read failures as misses, keep cache write failures non-fatal, and support refreshing an entry's TTL after ETag revalidation.
- Continue to prefer an explicitly configured static model catalog over an injected cache.
## Testing
- Cover injected cache hits, read and write failures, TTL refreshes, and end-to-end model selection.
- Verify that the file-backed implementation satisfies the new cache contract and does not serve expired entries before revalidation.
GitOrigin-RevId: eef957a317910b8502a0c20f8fc112f689d5e389
## What changed
- Remove the hidden `PairProgramming` and `Execute` variants from `ModeKind`.
- Delete their unused prompt templates and simplify mode handling to cover only `Default` and `Plan`.
GitOrigin-RevId: 8d24e3f562a09e3071b2decae275d2b3871a1e9c
## Why
The bundled system-skill cache is shared by every process using the same
`CODEX_HOME`. A service with bundled skills disabled must not remove cache
files that another service still uses.
## What changed
- Filter system skill roots from disabled services without deleting the shared
cache.
- Ensure bundled system skills are installed during loads when they are
enabled, including after a runtime configuration change.
## Testing
Add app-server coverage for preserving the cache across enabled and disabled
services and refreshing stale cached skills after runtime enablement.
GitOrigin-RevId: 1eca999b33b7be69873a40a91767fd26f0c2b457
## Why
Sequential dispatch lets a long-running request block unrelated health checks
and cleanup on the same connection.
## What changed
- Add `--concurrent-requests <COUNT>` for local and remote exec-server
connections, while retaining sequential dispatch when the option is omitted
or set to `1`.
- Preserve handshake ordering before enabling concurrent dispatch.
- Reserve separate capacity for status, signal, terminate, and close requests so
they remain responsive when ordinary request capacity is saturated.
- Drain queued client responses during disconnect and cancel outstanding
request tasks during connection shutdown.
## Testing
- Cover CLI parsing and concurrency-limit validation.
- Verify default sequential behavior, pipelined handshake ordering, concurrent
request progress, control-request responsiveness, and disconnect handling.
GitOrigin-RevId: 48e4b092e318204ed635543f01f9ee0e7df095fc
## What changed
- Add a hidden global `--psp` runtime flag and propagate it through TUI, exec,
app-server, remote-control, and in-process startup paths.
- Attach the `oai-chat-psp=true` cookie to first-party ChatGPT requests when
enabled, using a cached cookie-aware client with sensitive request logging
disabled.
- Keep the routing selection out of persistent configuration layers while
preserving it across config refreshes and agent role changes.
## Testing
- Cover global flag parsing, app-server propagation, config-layer isolation,
and preservation across config rebuilds and role changes.
GitOrigin-RevId: 05cdc61ffd7162d8e48fc1e166f4732113e5a816
## What changed
- Let `HttpClientFactory` carry additional ChatGPT cookies and share their store across cloned factories.
- Attach configured cookies when route-aware clients or explicitly opted-in clients enable the ChatGPT cookie store.
- Combine configured cookies with the current path-scoped Cloudflare cookies only for HTTPS ChatGPT hosts, while preserving sensitive-header metadata.
## Testing
- Cover cookie merging, path-scoped Cloudflare updates, host and scheme restrictions, sensitive values, and factory clone/equality behavior.
GitOrigin-RevId: 5baf745859e14db739fd62ab931720dbe90c04e0
## What changed
- Treat an MCP server as trusted for ChatGPT authentication when its origin
matches the configured HTTPS `chatgpt-staging.com` host or one of its
subdomains.
- Continue trusting the production ChatGPT origin and falling back to OAuth for
other transports and origins.
## Testing
- Cover both the staging apex domain and a staging subdomain.
GitOrigin-RevId: 1b1d06ab214141d39eb38d46de9dfb74e6fedb3e
## What changed
- Add provider-owned remote compaction capabilities for unsupported, v1, and v2 protocols.
- Mark Amazon Bedrock as v1-only so manual and automatic compaction use `/v1/responses/compact`, even when the v2 feature is enabled.
- Preserve v2 support for OpenAI and Azure Responses providers and local compaction for unsupported providers.
## Testing
- Cover Amazon Bedrock manual and automatic remote compaction, including endpoint selection, authentication headers, model selection, and reuse of the compacted summary.
GitOrigin-RevId: d58093178605d1f8155b93d9a161408177f6ac72
## What changed
Run the hash-pinned `openai/fence` action in `audit` mode before checkout in
the blob size policy job.
GitOrigin-RevId: 9312897dae76d93298008b2ce8a8767680fe34b8
## Why
Session IDs are derived from file stems and can repeat across projects. Keying
connector attribution only by session ID can therefore assign one session's
connectors to another session during a batched migration.
## What changed
- Preserve connector candidates by canonical source path until each imported
session is annotated, then aggregate them for detection summaries.
- Resolve project-backed MCP tool calls from both `CallMcpTool` server IDs and
`mcpDetails.serverName`, while requiring a match in project metadata.
## Testing
Add coverage for duplicate session stems, per-source connector results, and
both supported project-backed MCP tool-call formats.
GitOrigin-RevId: 10563d47c29673c07daa83be5f4e1a256740b901
## Why
Orchestrator skills with `allow_implicit_invocation: false` were still exposed in the model-visible skill catalog.
## What changed
- Hide explicit-only orchestrator skills from the prompt catalog while keeping them available for direct invocation.
- Include orchestrator resource-access metadata with injected instructions so `skills.read` can load referenced resources.
- Track visible and hidden discovery limits separately, and record discovery duration and resource-count metrics.
## Testing
Add an integration test that verifies an explicit-only orchestrator skill stays out of the prompt catalog, can be selected explicitly, and can read a referenced resource.
GitOrigin-RevId: 9cab5a61eb27c96bec9d1346c7dcdbc97376f7dc
## What changed
- Add a `features.token_budget.mode` setting with `thread` and `name` values.
- Default context-window metadata to the thread ID while allowing configurations to retain the agent name.
- Preserve the selected identity mode when applying model token-budget defaults and replaying locked configuration.
## Testing
- Cover configuration parsing and both identity formats in token-budget context messages.
GitOrigin-RevId: f6e6d2c49e09e70c0352e803bb65784ef77e5bee
## What changed
- Ignore symbolic links and other non-file, non-directory entries while copying a plugin into the store instead of rejecting the installation.
- Cover symlinked skill files and executables, verifying that the plugin installs without copying either link.
GitOrigin-RevId: da0cbbd95d12313afcaecdcb52824fc1922d0b74
## What changed
- Add the stable, default-enabled `features.view_image` flag.
- Omit the native `view_image` tool when the flag is disabled, including for fresh-context subagents and guardian reviewer turns.
- Keep unrelated execution and MCP tools available when the viewer is disabled.
## Testing
- Add app-server coverage for disabled viewer inheritance in fresh-context subagents and guardian reviewer turns.
GitOrigin-RevId: 42b1311010fbde9f064129b17b425820861c6d91
## Why
Cursor stores projectless chats under the reserved `empty-window` project, and
their transcripts may not contain embedded working-directory metadata.
## What changed
- Resolve `empty-window` sessions to the parent of the Cursor home directory,
including when that home is configured as a relative path.
- Record each created or appended session's working directory in its import
success result instead of inheriting the migration item's directory.
## Testing
- Cover projectless Cursor transcripts without embedded metadata.
- Cover working-directory resolution from a relative Cursor home.
GitOrigin-RevId: 21cc7ea6acab1ed80cc4016a881e07eca3d0abbd
## What changed
- Update the required pull request attribution to `Generated with [Codex](https://openai.com/codex/).`.
- Recognize the previous unlinked attribution when restoring saved context so it can be replaced without duplicating the attribution instructions.
## Testing
- Cover cold resume with both commit-only legacy instructions and the previous unlinked pull request attribution.
- Update app-server and MCP server assertions for the linked attribution.
GitOrigin-RevId: 8773546a6a59dab146905e50cf4434d53b9335cf
## Why
Trusting a directory enables project-local config, hooks, and exec policies, which can increase exposure to prompt injection. Require an explicit decision instead of automatically trusting projects whose trust level is unset.
## What changed
- Add a directory-trust step to TUI onboarding, with options to trust and continue or quit.
- Apply trust to the Git repository root when starting from a subdirectory, persist the decision, and reload config before continuing.
- Keep the prompt active and show the config error when trust cannot be persisted. Skip the prompt for remote workspaces and projects with an explicit trust level.
## Testing
- Cover trust-screen rendering, keyboard selection, persistence failures, and prompt visibility for explicit trust and Windows sandbox states.
GitOrigin-RevId: 850e6f54aec84d584cd161348ecdb1da89aaabd4
## What changed
- Return connector candidates found in detected external agent sessions, aggregating session counts by connector name.
- Resolve connector names from both cached plugin manifests and project MCP server metadata.
- Persist detected connector clues in the session import ledger and merge them into completed imports without case-insensitive duplicates.
## Testing
- Add coverage for project MCP metadata detection and for accumulating detected and imported connector candidates.
GitOrigin-RevId: 97a7dfe4a61599686b2222cddd9595acec8b4501
## Why
Image resize notices describe the history item immediately before them, so
remote compaction must not preserve or discard them independently of that item.
## What changed
- Group recognized `image_resize_notice` developer messages with their source
history items when filtering compacted history.
- Account for each group together when trimming or truncating history to a token
budget, preserving the notice only when its source is retained.
## Testing
- Extend the remote compaction v2 follow-up test to verify that a retained user
image keeps its resize notice, while notices for discarded items and
unrecognized developer notices are omitted.
GitOrigin-RevId: 82bb70a46eb68b07b4601e7fb345921d0789f442
## What changed
- Add `features.tool_registry.error_on_tool_collisions`, defaulting to `false`, to the TOML model, generated schema, and resolved runtime configuration.
- Treat `tool_registry` as structured configuration rather than a feature toggle, including in strict config validation and profile configuration.
- Preserve the resolved collision policy in session config lockfiles when it is enabled or explicitly configured.
## Testing
- Cover deserialization, strict validation, default and enabled resolution, feature materialization, and config lockfile serialization.
GitOrigin-RevId: 2c27109dcf2a3d1e51064cc60088703e36a0f85a
## What changed
- Add a storage-neutral `QueueStore` interface backed by a dedicated SQLite
database for ordered, thread-scoped user submissions.
- Support enqueueing, paginated listing, editing, deletion, and atomic
reordering, with a limit of 100 pending items per thread.
- Remove queued submissions when their thread is deleted.
## Testing
- Cover concurrent inserts, ordering and pagination, edits, deletion,
cross-thread isolation, queue limits, and thread cleanup.
GitOrigin-RevId: 0a8e95ac71137afc6f3a039d4b5fc8603110c891
## What changed
- Keep paginated resume, transcript, and fork workflows on bounded turn and item page requests, while preserving the legacy full-history fallback.
- Refill underfilled terminal scrollback from older pages and show a transcript shortcut when earlier history is unloaded or truncated by the row cap.
- Preserve a successfully created fork if its initial history hydration fails, keep side forks bounded, and retain the correct history boundary for nested forks and subagents.
## Testing
- Add coverage for scrollback refill and truncation notices, bounded paginated requests, fork hydration failures, nested forks, and inherited subagent history filtering.
GitOrigin-RevId: 10ba952ffb2f69717fbc71aa03cac7685f7dcf78
## What changed
- Load older transcript items in bounded pages as users scroll upward, and let `Home` continue fetching until the start of the thread.
- Show loading, partial-history, retry, and complete states in the transcript overlay while preserving the visible scroll position as pages are prepended.
- Use paginated history for resume, fork, retry, and prompt-edit flows, while falling back to legacy history requests when the app server does not support pagination.
## Testing
- Cover multi-page transcript loading, page-boundary prompt reconciliation, legacy-server fallback, bounded retry history, and transcript history-state rendering.
GitOrigin-RevId: cc6b764bc4185d079b2f68bedf235c1d08278dfb
## Why
Paginated sessions can have visible messages beyond the first item page, so
bounded hydration could leave resumed history or transcript previews empty and
the transcript viewer incomplete.
## What changed
- Scan paginated history until the configured rendered-row budget is filled,
while retaining the configured scrollback beyond the visible viewport.
- Fully hydrate transcripts opened from the resume picker and build previews
from bounded item pages, including when non-visible items precede messages.
- Stop pagination on repeated cursors and let `Esc` cancel an in-flight
transcript load.
- Replay restored token usage for cold metadata-only resumes when its stored
turn can be identified.
## Testing
- Add coverage for paginated and legacy transcript loading, bounded resume
hydration, repeated cursors, transcript cancellation, and resize reflow.
GitOrigin-RevId: 2ce69f7d5eae88e2b09a27c333b00b1830ff17bf
## Why
Paginated threads should not require the TUI to load and render their entire
history when a session is resumed.
## What changed
- Hydrate a bounded initial history page for paginated resumes and forks,
respecting the configured terminal reflow row budget.
- Load older transcript items in the background as users navigate upward, while
preserving the viewport, live output, and backtrack selection.
- Reconcile review-mode prompts across page boundaries and keep restored token
usage hidden until it can be attributed to a persisted turn.
## Testing
- Added coverage for paginated resume and fork hydration, older-page merging,
transcript prepending, backtrack behavior, review prompt filtering, and token
usage restoration.
GitOrigin-RevId: 42160b943a9f6a1ba0cfe4b195c24ddf410b747d
## What changed
- Generalize `CodexThread::try_start_turn_if_idle` to accept exported
`TurnInput` values, preserving user content and client IDs in the first model
request.
- Allow non-empty user input to start an idle turn in Plan mode while retaining
the Plan-mode rejection for automatic response items and empty user input.
- Process idle user input through the normal user-turn path so connector
selections are cleared, prompt telemetry is recorded, and referenced skill
instructions are included.
## Testing
- Cover idle user input in Default and Plan modes, response-item mailbox
merging, empty-input rejection, and skill instruction injection.
GitOrigin-RevId: 116667d160d4464b525d9e4c7e8cc05d23a94008
## What changed
- Move config-layer, user, system, plugin, extra, and repository skill-root
resolution from `core-skills` into the host skills extension.
- Keep the core loader focused on loading explicit `SkillRoot` values.
- Relocate and expand tests for root precedence, deduplication, repository
ancestry, plugin metadata, and concurrent probing.
GitOrigin-RevId: 3b95cf28101b8b4d64d54079d202154dad560aab
## Why
Reviews may use a different model from their parent turn. They need current
thread settings without inheriting defaults resolved for the parent model.
## What changed
- Build review configuration from the parent turn context while preserving
explicitly configured token-budget overrides.
- Use the review model's token-budget defaults and select a supported reasoning
effort when switching models.
- Carry the session's resolved service tier and current environment,
permissions, and approval settings into the review thread.
## Testing
Expanded review coverage for updated thread settings, model-specific
token-budget defaults, service tiers, and reasoning effort selection.
GitOrigin-RevId: cefa0f22cd9b9e28047d94e1d2d49ff83eded43e
## Why
Guardian reviews received sandbox retry reasons, but not the execution-policy reason that triggered an initial approval request.
## What changed
- Propagate approval and retry reasons through the tool approval flow and include the applicable reason in the Guardian prompt.
- Prefer a sandbox retry reason when both are available.
- Truncate approval context to 512 tokens while preserving the start and end of the reason.
## Testing
Added prompt and integration coverage for policy reasons, retry precedence, and truncation.
GitOrigin-RevId: 5b0f4e1c40b792b031bc8f7b31685a61d12b8d92
## Why
Amazon Bedrock supports hosted text web search, but it rejects the
`search_content_types` field used for multimodal search and does not support
external live or indexed web access.
## What changed
- Advertise hosted web search for Amazon Bedrock while marking external web
access as unsupported.
- Resolve unsupported live and indexed modes to cached search, or disable the
tool when cached search is prohibited by managed requirements.
- Normalize built-in and configured Bedrock model catalogs to text-only web
search, and retain the runtime provider in session configuration so turn
setup can apply its capabilities.
## Testing
- Cover cached fallback, managed-mode restrictions, text-only tool payloads,
provider capabilities, and catalog normalization.
GitOrigin-RevId: 310473849257401654388a4ebb42920e03aa3228
## What changed
- Replace the TUI directory-trust prompt with automatic trust for local projects whose trust level is unset. Keep explicit trust settings and remote workspaces unchanged.
- Persist trust for the resolved Git or configured project root, then reload configuration so project-local settings take effect.
- Fall back to an in-memory trust override and an embedded app server when the config update cannot be persisted.
## Testing
- Cover persisted and in-memory trust, custom project-root markers, and the working directories selected by resume and fork flows.
GitOrigin-RevId: 8fd51eb4cd88267073324bfd7dc4106a56d7c745
## What changed
- Remove the cached permission profile from `TurnContext` and derive the
effective profile, filesystem policy, network policy, and legacy sandbox
policy from `config.permissions`.
- Update sandbox consumers and tests to use the current turn configuration.
- Verify that role-based agent spawning reapplies runtime permissions and that
cold-resumed agents retain their disabled permission profile.
GitOrigin-RevId: d67c5d2bcbe6dc76f15b56defb485b155fd1f138
## Why
A repository can contain a tracked directory that Git implicitly treats as a bare
repository. Its configuration may select helpers such as `core.fsmonitor`, causing
Codex Git operations in that directory to execute repository-controlled code.
## What changed
- Pass `-c safe.bareRepository=explicit` to Codex-managed Git commands so they
reject implicitly discovered bare repositories.
- Continue to support repositories explicitly selected with `--git-dir` or
`GIT_DIR`.
## Testing
Add a regression test that clones a repository containing a tracked embedded Git
repository and verifies that guarded Git inspection rejects it without running
its configured filesystem monitor.
GitOrigin-RevId: 344b5bc1e0ffa94f2a1b788488aa653222da10f3
## What changed
- Move host skill discovery, snapshot caching, and configuration handling from
`codex-core-skills` to `codex-skills-extension`.
- Rename `SkillsService` and `SkillsLoadInput` to `HostSkillsService` and
`HostSkillsLoadInput` to distinguish the host implementation from other skill
providers.
- Keep shared loading primitives and skill outcome modeling in
`codex-core-skills`.
GitOrigin-RevId: d81a21791d2ef8c066e157e7e538b8cb7ee4c24b
