Files
codex/codex-rs/exec/tests/suite/agents_md.rs
T
Adam Perry @ OpenAI 236b50125d [codex] Load user instructions through an injected provider (#27101)
## Why

We want to remove implicit use of `$CODEX_HOME` from `codex-core` and
make embedders responsible for supplying user-level instructions. This
also ensures user instructions load when no primary environment is
selected.

## What changed

Stacked on #27415, which makes `codex exec` surface thread-scoped
runtime warnings.

- Added `UserInstructionsProvider` to `codex-extension-api`, with
absolute source attribution and recoverable loading warnings.
- Added `codex-home` with the filesystem-backed provider for
`AGENTS.override.md` and `AGENTS.md`, preserving precedence, fallback,
trimming, lossy UTF-8 handling, and the existing uncapped global
instruction size.
- Removed global instruction loading from `Config` and require
`ThreadManager` callers to inject a provider.
- Load provider instructions once for each fresh root runtime, including
runtimes without a primary environment. Running sessions retain their
snapshot, while child agents inherit the parent snapshot without
invoking the provider.
- Keep provider instructions separate while loading project `AGENTS.md`,
then assemble the model-visible instructions with the existing ordering,
source attribution, warning, and turn-context behavior.
- Wired the Codex home provider through the CLI, app server, MCP server,
core facade, and thread-manager sample.

## Validation

- `just test -p codex-home -p codex-extension-api`
- `just test -p codex-core agents_md`
- `just test -p codex-core guardian`
- `just test -p codex-app-server
thread_start_without_selected_environment_includes_only_global_instruction_source`
- `just test -p codex-exec warning`
- `just bazel-lock-check`
2026-06-11 19:28:47 +00:00

177 lines
6.0 KiB
Rust

#![allow(clippy::expect_used, clippy::unwrap_used)]
use core_test_support::responses;
use core_test_support::test_codex_exec::test_codex_exec;
use predicates::prelude::PredicateBooleanExt;
use predicates::str::contains;
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn exec_includes_workspace_agents_md_in_request() -> anyhow::Result<()> {
let test = test_codex_exec();
std::fs::write(test.cwd_path().join("AGENTS.md"), "workspace instructions")?;
let server = responses::start_mock_server().await;
let body = responses::sse(vec![
responses::ev_response_created("resp1"),
responses::ev_assistant_message("m1", "fixture hello"),
responses::ev_completed("resp1"),
]);
let response_mock = responses::mount_sse_once(&server, body).await;
test.cmd_with_server(&server)
.arg("--skip-git-repo-check")
.arg("tell me something")
.assert()
.success();
let user_messages = response_mock.single_request().message_input_texts("user");
assert!(
user_messages
.iter()
.any(|text| text.contains("workspace instructions")),
"request should include workspace AGENTS.md instructions: {user_messages:?}"
);
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn exec_prefers_workspace_agents_override_md() -> anyhow::Result<()> {
let test = test_codex_exec();
std::fs::write(test.cwd_path().join("AGENTS.md"), "base instructions")?;
std::fs::write(
test.cwd_path().join("AGENTS.override.md"),
"override instructions",
)?;
let server = responses::start_mock_server().await;
let body = responses::sse(vec![
responses::ev_response_created("resp1"),
responses::ev_assistant_message("m1", "fixture hello"),
responses::ev_completed("resp1"),
]);
let response_mock = responses::mount_sse_once(&server, body).await;
test.cmd_with_server(&server)
.arg("--skip-git-repo-check")
.arg("tell me something")
.assert()
.success();
let user_messages = response_mock.single_request().message_input_texts("user");
assert!(
user_messages
.iter()
.any(|text| text.contains("override instructions")),
"request should include AGENTS.override.md instructions: {user_messages:?}"
);
assert!(
user_messages
.iter()
.all(|text| !text.contains("base instructions")),
"request should exclude shadowed AGENTS.md instructions: {user_messages:?}"
);
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn exec_surfaces_project_instruction_loading_warnings() -> anyhow::Result<()> {
let test = test_codex_exec();
let project_agents_path = test.cwd_path().join("AGENTS.md");
std::fs::write(&project_agents_path, b"project\xFFinstructions")?;
let server = responses::start_mock_server().await;
let body = responses::sse(vec![
responses::ev_response_created("resp1"),
responses::ev_assistant_message("m1", "fixture hello"),
responses::ev_completed("resp1"),
]);
responses::mount_sse_once(&server, body).await;
test.cmd_with_server(&server)
.arg("--skip-git-repo-check")
.arg("tell me something")
.assert()
.success()
.stderr(contains("invalid UTF-8").and(contains(project_agents_path.display().to_string())));
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn exec_json_surfaces_project_instruction_loading_warnings() -> anyhow::Result<()> {
let test = test_codex_exec();
let project_agents_path = test.cwd_path().join("AGENTS.md");
std::fs::write(&project_agents_path, b"project\xFFinstructions")?;
let server = responses::start_mock_server().await;
let body = responses::sse(vec![
responses::ev_response_created("resp1"),
responses::ev_assistant_message("m1", "fixture hello"),
responses::ev_completed("resp1"),
]);
responses::mount_sse_once(&server, body).await;
let output = test
.cmd_with_server(&server)
.arg("--skip-git-repo-check")
.arg("--json")
.arg("tell me something")
.assert()
.success()
.get_output()
.stdout
.clone();
let events = String::from_utf8(output)?
.lines()
.map(serde_json::from_str::<serde_json::Value>)
.collect::<Result<Vec<_>, _>>()?;
assert!(
events.iter().any(|event| {
event["type"] == "item.completed"
&& event["item"]["type"] == "error"
&& event["item"]["message"].as_str().is_some_and(|message| {
message.contains("invalid UTF-8")
&& message.contains(project_agents_path.display().to_string().as_str())
})
}),
"expected a JSONL warning event; observed: {events:?}"
);
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn exec_surfaces_global_instruction_loading_warnings() -> anyhow::Result<()> {
let test = test_codex_exec();
let global_agents_path = test.home_path().join("AGENTS.md");
let global_agents_source_suffix = format!(
"{}{}AGENTS.md",
test.home_path()
.file_name()
.expect("temporary Codex home should have a file name")
.to_string_lossy(),
std::path::MAIN_SEPARATOR,
);
std::fs::write(&global_agents_path, b"global\xFFinstructions")?;
let server = responses::start_mock_server().await;
let body = responses::sse(vec![
responses::ev_response_created("resp1"),
responses::ev_assistant_message("m1", "fixture hello"),
responses::ev_completed("resp1"),
]);
responses::mount_sse_once(&server, body).await;
test.cmd_with_server(&server)
.arg("--skip-git-repo-check")
.arg("tell me something")
.assert()
.success()
.stderr(contains("invalid UTF-8").and(contains(global_agents_source_suffix)));
Ok(())
}