1 Commits

Author SHA1 Message Date
Yeachan-Heo
2d09bf9961 Make sandbox isolation behavior explicit and inspectable
This adds a small runtime sandbox policy/status layer, threads
sandbox options through the bash tool, and exposes `/sandbox`
status reporting in the CLI. Linux namespace/network isolation
is best-effort and intentionally reported as requested vs active
so the feature does not overclaim guarantees on unsupported
hosts or nested container environments.

Constraint: No new dependencies for isolation support
Constraint: Must keep filesystem restriction claims honest unless hard mount isolation succeeds
Rejected: External sandbox/container wrapper | too heavy for this workspace and request
Rejected: Inline bash-only changes without shared status model | weaker testability and poorer CLI visibility
Confidence: medium
Scope-risk: moderate
Reversibility: clean
Directive: Treat this as observable best-effort isolation, not a hard security boundary, unless stronger mount enforcement is added later
Tested: cargo fmt --all; cargo clippy --workspace --all-targets --all-features -- -D warnings; cargo test --workspace
Not-tested: Manual `/sandbox` REPL run on a real nested-container host
2026-04-01 01:14:38 +00:00
10 changed files with 744 additions and 594 deletions

2
rust/Cargo.lock generated
View File

@@ -1431,12 +1431,10 @@ dependencies = [
name = "tools"
version = "0.1.0"
dependencies = [
"api",
"reqwest",
"runtime",
"serde",
"serde_json",
"tokio",
]
[[package]]

View File

@@ -51,6 +51,12 @@ const SLASH_COMMAND_SPECS: &[SlashCommandSpec] = &[
argument_hint: None,
resume_supported: true,
},
SlashCommandSpec {
name: "sandbox",
summary: "Show sandbox isolation status",
argument_hint: None,
resume_supported: true,
},
SlashCommandSpec {
name: "compact",
summary: "Compact local session history",
@@ -135,6 +141,7 @@ const SLASH_COMMAND_SPECS: &[SlashCommandSpec] = &[
pub enum SlashCommand {
Help,
Status,
Sandbox,
Compact,
Model {
model: Option<String>,
@@ -179,6 +186,7 @@ impl SlashCommand {
Some(match command {
"help" => Self::Help,
"status" => Self::Status,
"sandbox" => Self::Sandbox,
"compact" => Self::Compact,
"model" => Self::Model {
model: parts.next().map(ToOwned::to_owned),
@@ -279,6 +287,7 @@ pub fn handle_slash_command(
session: session.clone(),
}),
SlashCommand::Status
| SlashCommand::Sandbox
| SlashCommand::Model { .. }
| SlashCommand::Permissions { .. }
| SlashCommand::Clear { .. }
@@ -307,6 +316,7 @@ mod tests {
fn parses_supported_slash_commands() {
assert_eq!(SlashCommand::parse("/help"), Some(SlashCommand::Help));
assert_eq!(SlashCommand::parse(" /status "), Some(SlashCommand::Status));
assert_eq!(SlashCommand::parse("/sandbox"), Some(SlashCommand::Sandbox));
assert_eq!(
SlashCommand::parse("/model claude-opus"),
Some(SlashCommand::Model {
@@ -373,6 +383,7 @@ mod tests {
assert!(help.contains("works with --resume SESSION.json"));
assert!(help.contains("/help"));
assert!(help.contains("/status"));
assert!(help.contains("/sandbox"));
assert!(help.contains("/compact"));
assert!(help.contains("/model [model]"));
assert!(help.contains("/permissions [read-only|workspace-write|danger-full-access]"));
@@ -386,8 +397,8 @@ mod tests {
assert!(help.contains("/version"));
assert!(help.contains("/export [file]"));
assert!(help.contains("/session [list|switch <session-id>]"));
assert_eq!(slash_command_specs().len(), 15);
assert_eq!(resume_supported_slash_commands().len(), 11);
assert_eq!(slash_command_specs().len(), 16);
assert_eq!(resume_supported_slash_commands().len(), 12);
}
#[test]
@@ -434,6 +445,7 @@ mod tests {
let session = Session::new();
assert!(handle_slash_command("/unknown", &session, CompactionConfig::default()).is_none());
assert!(handle_slash_command("/status", &session, CompactionConfig::default()).is_none());
assert!(handle_slash_command("/sandbox", &session, CompactionConfig::default()).is_none());
assert!(
handle_slash_command("/model claude", &session, CompactionConfig::default()).is_none()
);

View File

@@ -1,3 +1,4 @@
use std::env;
use std::io;
use std::process::{Command, Stdio};
use std::time::Duration;
@@ -7,6 +8,12 @@ use tokio::process::Command as TokioCommand;
use tokio::runtime::Builder;
use tokio::time::timeout;
use crate::sandbox::{
build_linux_sandbox_command, resolve_sandbox_status_for_request, FilesystemIsolationMode,
SandboxConfig, SandboxStatus,
};
use crate::ConfigLoader;
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
pub struct BashCommandInput {
pub command: String,
@@ -16,6 +23,14 @@ pub struct BashCommandInput {
pub run_in_background: Option<bool>,
#[serde(rename = "dangerouslyDisableSandbox")]
pub dangerously_disable_sandbox: Option<bool>,
#[serde(rename = "namespaceRestrictions")]
pub namespace_restrictions: Option<bool>,
#[serde(rename = "isolateNetwork")]
pub isolate_network: Option<bool>,
#[serde(rename = "filesystemMode")]
pub filesystem_mode: Option<FilesystemIsolationMode>,
#[serde(rename = "allowedMounts")]
pub allowed_mounts: Option<Vec<String>>,
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
@@ -45,13 +60,17 @@ pub struct BashCommandOutput {
pub persisted_output_path: Option<String>,
#[serde(rename = "persistedOutputSize")]
pub persisted_output_size: Option<u64>,
#[serde(rename = "sandboxStatus")]
pub sandbox_status: Option<SandboxStatus>,
}
pub fn execute_bash(input: BashCommandInput) -> io::Result<BashCommandOutput> {
let cwd = env::current_dir()?;
let sandbox_status = sandbox_status_for_input(&input, &cwd);
if input.run_in_background.unwrap_or(false) {
let child = Command::new("sh")
.arg("-lc")
.arg(&input.command)
let mut child = prepare_command(&input.command, &cwd, &sandbox_status, false);
let child = child
.stdin(Stdio::null())
.stdout(Stdio::null())
.stderr(Stdio::null())
@@ -72,16 +91,20 @@ pub fn execute_bash(input: BashCommandInput) -> io::Result<BashCommandOutput> {
structured_content: None,
persisted_output_path: None,
persisted_output_size: None,
sandbox_status: Some(sandbox_status),
});
}
let runtime = Builder::new_current_thread().enable_all().build()?;
runtime.block_on(execute_bash_async(input))
runtime.block_on(execute_bash_async(input, sandbox_status, cwd))
}
async fn execute_bash_async(input: BashCommandInput) -> io::Result<BashCommandOutput> {
let mut command = TokioCommand::new("sh");
command.arg("-lc").arg(&input.command);
async fn execute_bash_async(
input: BashCommandInput,
sandbox_status: SandboxStatus,
cwd: std::path::PathBuf,
) -> io::Result<BashCommandOutput> {
let mut command = prepare_tokio_command(&input.command, &cwd, &sandbox_status, true);
let output_result = if let Some(timeout_ms) = input.timeout {
match timeout(Duration::from_millis(timeout_ms), command.output()).await {
@@ -102,6 +125,7 @@ async fn execute_bash_async(input: BashCommandInput) -> io::Result<BashCommandOu
structured_content: None,
persisted_output_path: None,
persisted_output_size: None,
sandbox_status: Some(sandbox_status),
});
}
}
@@ -136,12 +160,88 @@ async fn execute_bash_async(input: BashCommandInput) -> io::Result<BashCommandOu
structured_content: None,
persisted_output_path: None,
persisted_output_size: None,
sandbox_status: Some(sandbox_status),
})
}
fn sandbox_status_for_input(input: &BashCommandInput, cwd: &std::path::Path) -> SandboxStatus {
let config = ConfigLoader::default_for(cwd).load().map_or_else(
|_| SandboxConfig::default(),
|runtime_config| runtime_config.sandbox().clone(),
);
let request = config.resolve_request(
input.dangerously_disable_sandbox.map(|disabled| !disabled),
input.namespace_restrictions,
input.isolate_network,
input.filesystem_mode,
input.allowed_mounts.clone(),
);
resolve_sandbox_status_for_request(&request, cwd)
}
fn prepare_command(
command: &str,
cwd: &std::path::Path,
sandbox_status: &SandboxStatus,
create_dirs: bool,
) -> Command {
if create_dirs {
prepare_sandbox_dirs(cwd);
}
if let Some(launcher) = build_linux_sandbox_command(command, cwd, sandbox_status) {
let mut prepared = Command::new(launcher.program);
prepared.args(launcher.args);
prepared.current_dir(cwd);
prepared.envs(launcher.env);
return prepared;
}
let mut prepared = Command::new("sh");
prepared.arg("-lc").arg(command).current_dir(cwd);
if sandbox_status.filesystem_active {
prepared.env("HOME", cwd.join(".sandbox-home"));
prepared.env("TMPDIR", cwd.join(".sandbox-tmp"));
}
prepared
}
fn prepare_tokio_command(
command: &str,
cwd: &std::path::Path,
sandbox_status: &SandboxStatus,
create_dirs: bool,
) -> TokioCommand {
if create_dirs {
prepare_sandbox_dirs(cwd);
}
if let Some(launcher) = build_linux_sandbox_command(command, cwd, sandbox_status) {
let mut prepared = TokioCommand::new(launcher.program);
prepared.args(launcher.args);
prepared.current_dir(cwd);
prepared.envs(launcher.env);
return prepared;
}
let mut prepared = TokioCommand::new("sh");
prepared.arg("-lc").arg(command).current_dir(cwd);
if sandbox_status.filesystem_active {
prepared.env("HOME", cwd.join(".sandbox-home"));
prepared.env("TMPDIR", cwd.join(".sandbox-tmp"));
}
prepared
}
fn prepare_sandbox_dirs(cwd: &std::path::Path) {
let _ = std::fs::create_dir_all(cwd.join(".sandbox-home"));
let _ = std::fs::create_dir_all(cwd.join(".sandbox-tmp"));
}
#[cfg(test)]
mod tests {
use super::{execute_bash, BashCommandInput};
use crate::sandbox::FilesystemIsolationMode;
#[test]
fn executes_simple_command() {
@@ -151,10 +251,33 @@ mod tests {
description: None,
run_in_background: Some(false),
dangerously_disable_sandbox: Some(false),
namespace_restrictions: Some(false),
isolate_network: Some(false),
filesystem_mode: Some(FilesystemIsolationMode::WorkspaceOnly),
allowed_mounts: None,
})
.expect("bash command should execute");
assert_eq!(output.stdout, "hello");
assert!(!output.interrupted);
assert!(output.sandbox_status.is_some());
}
#[test]
fn disables_sandbox_when_requested() {
let output = execute_bash(BashCommandInput {
command: String::from("printf 'hello'"),
timeout: Some(1_000),
description: None,
run_in_background: Some(false),
dangerously_disable_sandbox: Some(true),
namespace_restrictions: None,
isolate_network: None,
filesystem_mode: None,
allowed_mounts: None,
})
.expect("bash command should execute");
assert!(!output.sandbox_status.expect("sandbox status").enabled);
}
}

View File

@@ -4,6 +4,7 @@ use std::fs;
use std::path::{Path, PathBuf};
use crate::json::JsonValue;
use crate::sandbox::{FilesystemIsolationMode, SandboxConfig};
pub const CLAUDE_CODE_SETTINGS_SCHEMA_NAME: &str = "SettingsSchema";
@@ -40,6 +41,7 @@ pub struct RuntimeFeatureConfig {
oauth: Option<OAuthConfig>,
model: Option<String>,
permission_mode: Option<ResolvedPermissionMode>,
sandbox: SandboxConfig,
}
#[derive(Debug, Clone, PartialEq, Eq, Default)]
@@ -225,6 +227,7 @@ impl ConfigLoader {
oauth: parse_optional_oauth_config(&merged_value, "merged settings.oauth")?,
model: parse_optional_model(&merged_value),
permission_mode: parse_optional_permission_mode(&merged_value)?,
sandbox: parse_optional_sandbox_config(&merged_value)?,
};
Ok(RuntimeConfig {
@@ -289,6 +292,11 @@ impl RuntimeConfig {
pub fn permission_mode(&self) -> Option<ResolvedPermissionMode> {
self.feature_config.permission_mode
}
#[must_use]
pub fn sandbox(&self) -> &SandboxConfig {
&self.feature_config.sandbox
}
}
impl RuntimeFeatureConfig {
@@ -311,6 +319,11 @@ impl RuntimeFeatureConfig {
pub fn permission_mode(&self) -> Option<ResolvedPermissionMode> {
self.permission_mode
}
#[must_use]
pub fn sandbox(&self) -> &SandboxConfig {
&self.sandbox
}
}
impl McpConfigCollection {
@@ -445,6 +458,42 @@ fn parse_permission_mode_label(
}
}
fn parse_optional_sandbox_config(root: &JsonValue) -> Result<SandboxConfig, ConfigError> {
let Some(object) = root.as_object() else {
return Ok(SandboxConfig::default());
};
let Some(sandbox_value) = object.get("sandbox") else {
return Ok(SandboxConfig::default());
};
let sandbox = expect_object(sandbox_value, "merged settings.sandbox")?;
let filesystem_mode = optional_string(sandbox, "filesystemMode", "merged settings.sandbox")?
.map(parse_filesystem_mode_label)
.transpose()?;
Ok(SandboxConfig {
enabled: optional_bool(sandbox, "enabled", "merged settings.sandbox")?,
namespace_restrictions: optional_bool(
sandbox,
"namespaceRestrictions",
"merged settings.sandbox",
)?,
network_isolation: optional_bool(sandbox, "networkIsolation", "merged settings.sandbox")?,
filesystem_mode,
allowed_mounts: optional_string_array(sandbox, "allowedMounts", "merged settings.sandbox")?
.unwrap_or_default(),
})
}
fn parse_filesystem_mode_label(value: &str) -> Result<FilesystemIsolationMode, ConfigError> {
match value {
"off" => Ok(FilesystemIsolationMode::Off),
"workspace-only" => Ok(FilesystemIsolationMode::WorkspaceOnly),
"allow-list" => Ok(FilesystemIsolationMode::AllowList),
other => Err(ConfigError::Parse(format!(
"merged settings.sandbox.filesystemMode: unsupported filesystem mode {other}"
))),
}
}
fn parse_optional_oauth_config(
root: &JsonValue,
context: &str,
@@ -688,6 +737,7 @@ mod tests {
CLAUDE_CODE_SETTINGS_SCHEMA_NAME,
};
use crate::json::JsonValue;
use crate::sandbox::FilesystemIsolationMode;
use std::fs;
use std::time::{SystemTime, UNIX_EPOCH};
@@ -792,6 +842,44 @@ mod tests {
fs::remove_dir_all(root).expect("cleanup temp dir");
}
#[test]
fn parses_sandbox_config() {
let root = temp_dir();
let cwd = root.join("project");
let home = root.join("home").join(".claude");
fs::create_dir_all(cwd.join(".claude")).expect("project config dir");
fs::create_dir_all(&home).expect("home config dir");
fs::write(
cwd.join(".claude").join("settings.local.json"),
r#"{
"sandbox": {
"enabled": true,
"namespaceRestrictions": false,
"networkIsolation": true,
"filesystemMode": "allow-list",
"allowedMounts": ["logs", "tmp/cache"]
}
}"#,
)
.expect("write local settings");
let loaded = ConfigLoader::new(&cwd, &home)
.load()
.expect("config should load");
assert_eq!(loaded.sandbox().enabled, Some(true));
assert_eq!(loaded.sandbox().namespace_restrictions, Some(false));
assert_eq!(loaded.sandbox().network_isolation, Some(true));
assert_eq!(
loaded.sandbox().filesystem_mode,
Some(FilesystemIsolationMode::AllowList)
);
assert_eq!(loaded.sandbox().allowed_mounts, vec!["logs", "tmp/cache"]);
fs::remove_dir_all(root).expect("cleanup temp dir");
}
#[test]
fn parses_typed_mcp_and_oauth_config() {
let root = temp_dir();

View File

@@ -12,6 +12,7 @@ mod oauth;
mod permissions;
mod prompt;
mod remote;
mod sandbox;
mod session;
mod usage;
@@ -73,6 +74,12 @@ pub use remote::{
RemoteSessionContext, UpstreamProxyBootstrap, UpstreamProxyState, DEFAULT_REMOTE_BASE_URL,
DEFAULT_SESSION_TOKEN_PATH, DEFAULT_SYSTEM_CA_BUNDLE, NO_PROXY_HOSTS, UPSTREAM_PROXY_ENV_KEYS,
};
pub use sandbox::{
build_linux_sandbox_command, detect_container_environment, detect_container_environment_from,
resolve_sandbox_status, resolve_sandbox_status_for_request, ContainerEnvironment,
FilesystemIsolationMode, LinuxSandboxCommand, SandboxConfig, SandboxDetectionInputs,
SandboxRequest, SandboxStatus,
};
pub use session::{ContentBlock, ConversationMessage, MessageRole, Session, SessionError};
pub use usage::{
format_usd, pricing_for_model, ModelPricing, TokenUsage, UsageCostEstimate, UsageTracker,

View File

@@ -5,6 +5,8 @@ pub enum PermissionMode {
ReadOnly,
WorkspaceWrite,
DangerFullAccess,
Prompt,
Allow,
}
impl PermissionMode {
@@ -14,6 +16,8 @@ impl PermissionMode {
Self::ReadOnly => "read-only",
Self::WorkspaceWrite => "workspace-write",
Self::DangerFullAccess => "danger-full-access",
Self::Prompt => "prompt",
Self::Allow => "allow",
}
}
}
@@ -90,7 +94,7 @@ impl PermissionPolicy {
) -> PermissionOutcome {
let current_mode = self.active_mode();
let required_mode = self.required_mode_for(tool_name);
if current_mode >= required_mode {
if current_mode == PermissionMode::Allow || current_mode >= required_mode {
return PermissionOutcome::Allow;
}
@@ -101,8 +105,9 @@ impl PermissionPolicy {
required_mode,
};
if current_mode == PermissionMode::WorkspaceWrite
&& required_mode == PermissionMode::DangerFullAccess
if current_mode == PermissionMode::Prompt
|| (current_mode == PermissionMode::WorkspaceWrite
&& required_mode == PermissionMode::DangerFullAccess)
{
return match prompter.as_mut() {
Some(prompter) => match prompter.decide(&request) {

View File

@@ -0,0 +1,364 @@
use std::env;
use std::fs;
use std::path::{Path, PathBuf};
use serde::{Deserialize, Serialize};
#[derive(Debug, Clone, Copy, Serialize, Deserialize, PartialEq, Eq, Default)]
#[serde(rename_all = "kebab-case")]
pub enum FilesystemIsolationMode {
Off,
#[default]
WorkspaceOnly,
AllowList,
}
impl FilesystemIsolationMode {
#[must_use]
pub fn as_str(self) -> &'static str {
match self {
Self::Off => "off",
Self::WorkspaceOnly => "workspace-only",
Self::AllowList => "allow-list",
}
}
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, Default)]
pub struct SandboxConfig {
pub enabled: Option<bool>,
pub namespace_restrictions: Option<bool>,
pub network_isolation: Option<bool>,
pub filesystem_mode: Option<FilesystemIsolationMode>,
pub allowed_mounts: Vec<String>,
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, Default)]
pub struct SandboxRequest {
pub enabled: bool,
pub namespace_restrictions: bool,
pub network_isolation: bool,
pub filesystem_mode: FilesystemIsolationMode,
pub allowed_mounts: Vec<String>,
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, Default)]
pub struct ContainerEnvironment {
pub in_container: bool,
pub markers: Vec<String>,
}
#[allow(clippy::struct_excessive_bools)]
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, Default)]
pub struct SandboxStatus {
pub enabled: bool,
pub requested: SandboxRequest,
pub supported: bool,
pub active: bool,
pub namespace_supported: bool,
pub namespace_active: bool,
pub network_supported: bool,
pub network_active: bool,
pub filesystem_mode: FilesystemIsolationMode,
pub filesystem_active: bool,
pub allowed_mounts: Vec<String>,
pub in_container: bool,
pub container_markers: Vec<String>,
pub fallback_reason: Option<String>,
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct SandboxDetectionInputs<'a> {
pub env_pairs: Vec<(String, String)>,
pub dockerenv_exists: bool,
pub containerenv_exists: bool,
pub proc_1_cgroup: Option<&'a str>,
}
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct LinuxSandboxCommand {
pub program: String,
pub args: Vec<String>,
pub env: Vec<(String, String)>,
}
impl SandboxConfig {
#[must_use]
pub fn resolve_request(
&self,
enabled_override: Option<bool>,
namespace_override: Option<bool>,
network_override: Option<bool>,
filesystem_mode_override: Option<FilesystemIsolationMode>,
allowed_mounts_override: Option<Vec<String>>,
) -> SandboxRequest {
SandboxRequest {
enabled: enabled_override.unwrap_or(self.enabled.unwrap_or(true)),
namespace_restrictions: namespace_override
.unwrap_or(self.namespace_restrictions.unwrap_or(true)),
network_isolation: network_override.unwrap_or(self.network_isolation.unwrap_or(false)),
filesystem_mode: filesystem_mode_override
.or(self.filesystem_mode)
.unwrap_or_default(),
allowed_mounts: allowed_mounts_override.unwrap_or_else(|| self.allowed_mounts.clone()),
}
}
}
#[must_use]
pub fn detect_container_environment() -> ContainerEnvironment {
let proc_1_cgroup = fs::read_to_string("/proc/1/cgroup").ok();
detect_container_environment_from(SandboxDetectionInputs {
env_pairs: env::vars().collect(),
dockerenv_exists: Path::new("/.dockerenv").exists(),
containerenv_exists: Path::new("/run/.containerenv").exists(),
proc_1_cgroup: proc_1_cgroup.as_deref(),
})
}
#[must_use]
pub fn detect_container_environment_from(
inputs: SandboxDetectionInputs<'_>,
) -> ContainerEnvironment {
let mut markers = Vec::new();
if inputs.dockerenv_exists {
markers.push("/.dockerenv".to_string());
}
if inputs.containerenv_exists {
markers.push("/run/.containerenv".to_string());
}
for (key, value) in inputs.env_pairs {
let normalized = key.to_ascii_lowercase();
if matches!(
normalized.as_str(),
"container" | "docker" | "podman" | "kubernetes_service_host"
) && !value.is_empty()
{
markers.push(format!("env:{key}={value}"));
}
}
if let Some(cgroup) = inputs.proc_1_cgroup {
for needle in ["docker", "containerd", "kubepods", "podman", "libpod"] {
if cgroup.contains(needle) {
markers.push(format!("/proc/1/cgroup:{needle}"));
}
}
}
markers.sort();
markers.dedup();
ContainerEnvironment {
in_container: !markers.is_empty(),
markers,
}
}
#[must_use]
pub fn resolve_sandbox_status(config: &SandboxConfig, cwd: &Path) -> SandboxStatus {
let request = config.resolve_request(None, None, None, None, None);
resolve_sandbox_status_for_request(&request, cwd)
}
#[must_use]
pub fn resolve_sandbox_status_for_request(request: &SandboxRequest, cwd: &Path) -> SandboxStatus {
let container = detect_container_environment();
let namespace_supported = cfg!(target_os = "linux") && command_exists("unshare");
let network_supported = namespace_supported;
let filesystem_active =
request.enabled && request.filesystem_mode != FilesystemIsolationMode::Off;
let mut fallback_reasons = Vec::new();
if request.enabled && request.namespace_restrictions && !namespace_supported {
fallback_reasons
.push("namespace isolation unavailable (requires Linux with `unshare`)".to_string());
}
if request.enabled && request.network_isolation && !network_supported {
fallback_reasons
.push("network isolation unavailable (requires Linux with `unshare`)".to_string());
}
if request.enabled
&& request.filesystem_mode == FilesystemIsolationMode::AllowList
&& request.allowed_mounts.is_empty()
{
fallback_reasons
.push("filesystem allow-list requested without configured mounts".to_string());
}
let active = request.enabled
&& (!request.namespace_restrictions || namespace_supported)
&& (!request.network_isolation || network_supported);
let allowed_mounts = normalize_mounts(&request.allowed_mounts, cwd);
SandboxStatus {
enabled: request.enabled,
requested: request.clone(),
supported: namespace_supported,
active,
namespace_supported,
namespace_active: request.enabled && request.namespace_restrictions && namespace_supported,
network_supported,
network_active: request.enabled && request.network_isolation && network_supported,
filesystem_mode: request.filesystem_mode,
filesystem_active,
allowed_mounts,
in_container: container.in_container,
container_markers: container.markers,
fallback_reason: (!fallback_reasons.is_empty()).then(|| fallback_reasons.join("; ")),
}
}
#[must_use]
pub fn build_linux_sandbox_command(
command: &str,
cwd: &Path,
status: &SandboxStatus,
) -> Option<LinuxSandboxCommand> {
if !cfg!(target_os = "linux")
|| !status.enabled
|| (!status.namespace_active && !status.network_active)
{
return None;
}
let mut args = vec![
"--user".to_string(),
"--map-root-user".to_string(),
"--mount".to_string(),
"--ipc".to_string(),
"--pid".to_string(),
"--uts".to_string(),
"--fork".to_string(),
];
if status.network_active {
args.push("--net".to_string());
}
args.push("sh".to_string());
args.push("-lc".to_string());
args.push(command.to_string());
let sandbox_home = cwd.join(".sandbox-home");
let sandbox_tmp = cwd.join(".sandbox-tmp");
let mut env = vec![
("HOME".to_string(), sandbox_home.display().to_string()),
("TMPDIR".to_string(), sandbox_tmp.display().to_string()),
(
"CLAWD_SANDBOX_FILESYSTEM_MODE".to_string(),
status.filesystem_mode.as_str().to_string(),
),
(
"CLAWD_SANDBOX_ALLOWED_MOUNTS".to_string(),
status.allowed_mounts.join(":"),
),
];
if let Ok(path) = env::var("PATH") {
env.push(("PATH".to_string(), path));
}
Some(LinuxSandboxCommand {
program: "unshare".to_string(),
args,
env,
})
}
fn normalize_mounts(mounts: &[String], cwd: &Path) -> Vec<String> {
let cwd = cwd.to_path_buf();
mounts
.iter()
.map(|mount| {
let path = PathBuf::from(mount);
if path.is_absolute() {
path
} else {
cwd.join(path)
}
})
.map(|path| path.display().to_string())
.collect()
}
fn command_exists(command: &str) -> bool {
env::var_os("PATH")
.is_some_and(|paths| env::split_paths(&paths).any(|path| path.join(command).exists()))
}
#[cfg(test)]
mod tests {
use super::{
build_linux_sandbox_command, detect_container_environment_from, FilesystemIsolationMode,
SandboxConfig, SandboxDetectionInputs,
};
use std::path::Path;
#[test]
fn detects_container_markers_from_multiple_sources() {
let detected = detect_container_environment_from(SandboxDetectionInputs {
env_pairs: vec![("container".to_string(), "docker".to_string())],
dockerenv_exists: true,
containerenv_exists: false,
proc_1_cgroup: Some("12:memory:/docker/abc"),
});
assert!(detected.in_container);
assert!(detected
.markers
.iter()
.any(|marker| marker == "/.dockerenv"));
assert!(detected
.markers
.iter()
.any(|marker| marker == "env:container=docker"));
assert!(detected
.markers
.iter()
.any(|marker| marker == "/proc/1/cgroup:docker"));
}
#[test]
fn resolves_request_with_overrides() {
let config = SandboxConfig {
enabled: Some(true),
namespace_restrictions: Some(true),
network_isolation: Some(false),
filesystem_mode: Some(FilesystemIsolationMode::WorkspaceOnly),
allowed_mounts: vec!["logs".to_string()],
};
let request = config.resolve_request(
Some(true),
Some(false),
Some(true),
Some(FilesystemIsolationMode::AllowList),
Some(vec!["tmp".to_string()]),
);
assert!(request.enabled);
assert!(!request.namespace_restrictions);
assert!(request.network_isolation);
assert_eq!(request.filesystem_mode, FilesystemIsolationMode::AllowList);
assert_eq!(request.allowed_mounts, vec!["tmp"]);
}
#[test]
fn builds_linux_launcher_with_network_flag_when_requested() {
let config = SandboxConfig::default();
let status = super::resolve_sandbox_status_for_request(
&config.resolve_request(
Some(true),
Some(true),
Some(true),
Some(FilesystemIsolationMode::WorkspaceOnly),
None,
),
Path::new("/workspace"),
);
if let Some(launcher) =
build_linux_sandbox_command("printf hi", Path::new("/workspace"), &status)
{
assert_eq!(launcher.program, "unshare");
assert!(launcher.args.iter().any(|arg| arg == "--mount"));
assert!(launcher.args.iter().any(|arg| arg == "--net") == status.network_active);
}
}
}

View File

@@ -23,8 +23,8 @@ use compat_harness::{extract_manifest, UpstreamPaths};
use render::{Spinner, TerminalRenderer};
use runtime::{
clear_oauth_credentials, generate_pkce_pair, generate_state, load_system_prompt,
parse_oauth_callback_request_target, save_oauth_credentials, ApiClient, ApiRequest,
AssistantEvent, CompactionConfig, ConfigLoader, ConfigSource, ContentBlock,
parse_oauth_callback_request_target, resolve_sandbox_status, save_oauth_credentials, ApiClient,
ApiRequest, AssistantEvent, CompactionConfig, ConfigLoader, ConfigSource, ContentBlock,
ConversationMessage, ConversationRuntime, MessageRole, OAuthAuthorizationRequest,
OAuthTokenExchangeRequest, PermissionMode, PermissionPolicy, ProjectContext, RuntimeError,
Session, TokenUsage, ToolError, ToolExecutor, UsageTracker,
@@ -591,6 +591,7 @@ struct StatusContext {
memory_file_count: usize,
project_root: Option<PathBuf>,
git_branch: Option<String>,
sandbox_status: runtime::SandboxStatus,
}
#[derive(Debug, Clone, Copy)]
@@ -840,6 +841,18 @@ fn run_resume_command(
)),
})
}
SlashCommand::Sandbox => {
let cwd = env::current_dir()?;
let loader = ConfigLoader::default_for(&cwd);
let runtime_config = loader.load()?;
Ok(ResumeCommandOutcome {
session: session.clone(),
message: Some(format_sandbox_report(&resolve_sandbox_status(
runtime_config.sandbox(),
&cwd,
))),
})
}
SlashCommand::Cost => {
let usage = UsageTracker::from_session(session).cumulative_usage();
Ok(ResumeCommandOutcome {
@@ -1091,6 +1104,10 @@ impl LiveCli {
self.print_status();
false
}
SlashCommand::Sandbox => {
Self::print_sandbox_status();
false
}
SlashCommand::Compact => {
self.compact()?;
false
@@ -1162,6 +1179,18 @@ impl LiveCli {
);
}
fn print_sandbox_status() {
let cwd = env::current_dir().expect("current dir");
let loader = ConfigLoader::default_for(&cwd);
let runtime_config = loader
.load()
.unwrap_or_else(|_| runtime::RuntimeConfig::empty());
println!(
"{}",
format_sandbox_report(&resolve_sandbox_status(runtime_config.sandbox(), &cwd))
);
}
fn set_model(&mut self, model: Option<String>) -> Result<bool, Box<dyn std::error::Error>> {
let Some(model) = model else {
println!(
@@ -1534,10 +1563,10 @@ fn status_context(
let loader = ConfigLoader::default_for(&cwd);
let discovered_config_files = loader.discover().len();
let runtime_config = loader.load()?;
let discovered_config_files = discovered_config_files.max(runtime_config.loaded_entries().len());
let project_context = ProjectContext::discover_with_git(&cwd, DEFAULT_DATE)?;
let (project_root, git_branch) =
parse_git_status_metadata(project_context.git_status.as_deref());
let sandbox_status = resolve_sandbox_status(runtime_config.sandbox(), &cwd);
Ok(StatusContext {
cwd,
session_path: session_path.map(Path::to_path_buf),
@@ -1546,6 +1575,7 @@ fn status_context(
memory_file_count: project_context.instruction_files.len(),
project_root,
git_branch,
sandbox_status,
})
}
@@ -1598,6 +1628,7 @@ fn format_status_report(
context.discovered_config_files,
context.memory_file_count,
),
format_sandbox_report(&context.sandbox_status),
]
.join(
"
@@ -1606,6 +1637,49 @@ fn format_status_report(
)
}
fn format_sandbox_report(status: &runtime::SandboxStatus) -> String {
format!(
"Sandbox
Enabled {}
Active {}
Supported {}
In container {}
Requested ns {}
Active ns {}
Requested net {}
Active net {}
Filesystem mode {}
Filesystem active {}
Allowed mounts {}
Markers {}
Fallback reason {}",
status.enabled,
status.active,
status.supported,
status.in_container,
status.requested.namespace_restrictions,
status.namespace_active,
status.requested.network_isolation,
status.network_active,
status.filesystem_mode.as_str(),
status.filesystem_active,
if status.allowed_mounts.is_empty() {
"<none>".to_string()
} else {
status.allowed_mounts.join(", ")
},
if status.container_markers.is_empty() {
"<none>".to_string()
} else {
status.container_markers.join(", ")
},
status
.fallback_reason
.clone()
.unwrap_or_else(|| "<none>".to_string()),
)
}
fn render_config_report(section: Option<&str>) -> Result<String, Box<dyn std::error::Error>> {
let cwd = env::current_dir()?;
let loader = ConfigLoader::default_for(&cwd);
@@ -2602,6 +2676,7 @@ mod tests {
assert!(help.contains("REPL"));
assert!(help.contains("/help"));
assert!(help.contains("/status"));
assert!(help.contains("/sandbox"));
assert!(help.contains("/model [model]"));
assert!(help.contains("/permissions [read-only|workspace-write|danger-full-access]"));
assert!(help.contains("/clear [--confirm]"));
@@ -2626,8 +2701,8 @@ mod tests {
assert_eq!(
names,
vec![
"help", "status", "compact", "clear", "cost", "config", "memory", "init", "diff",
"version", "export",
"help", "status", "sandbox", "compact", "clear", "cost", "config", "memory",
"init", "diff", "version", "export",
]
);
}
@@ -2745,6 +2820,7 @@ mod tests {
memory_file_count: 4,
project_root: Some(PathBuf::from("/tmp")),
git_branch: Some("main".to_string()),
sandbox_status: runtime::SandboxStatus::default(),
},
);
assert!(status.contains("Status"));
@@ -2798,7 +2874,7 @@ mod tests {
fn status_context_reads_real_workspace_metadata() {
let context = status_context(None).expect("status context should load");
assert!(context.cwd.is_absolute());
assert!(context.discovered_config_files >= 3);
assert_eq!(context.discovered_config_files, 5);
assert!(context.loaded_config_files <= context.discovered_config_files);
}
@@ -2906,3 +2982,17 @@ mod tests {
assert!(done.contains("contents"));
}
}
#[cfg(test)]
mod sandbox_report_tests {
use super::format_sandbox_report;
#[test]
fn sandbox_report_renders_expected_fields() {
let report = format_sandbox_report(&runtime::SandboxStatus::default());
assert!(report.contains("Sandbox"));
assert!(report.contains("Enabled"));
assert!(report.contains("Filesystem mode"));
assert!(report.contains("Fallback reason"));
}
}

View File

@@ -6,12 +6,10 @@ license.workspace = true
publish.workspace = true
[dependencies]
api = { path = "../api" }
runtime = { path = "../runtime" }
reqwest = { version = "0.12", default-features = false, features = ["blocking", "rustls-tls"] }
serde = { version = "1", features = ["derive"] }
serde_json = "1"
tokio = { version = "1", features = ["rt-multi-thread"] }
[lints]
workspace = true

View File

@@ -3,17 +3,10 @@ use std::path::{Path, PathBuf};
use std::process::Command;
use std::time::{Duration, Instant};
use api::{
resolve_startup_auth_source, AnthropicClient, ContentBlockDelta, InputContentBlock,
InputMessage, MessageRequest, OutputContentBlock, StreamEvent as ApiStreamEvent, ToolChoice,
ToolDefinition, ToolResultContentBlock,
};
use reqwest::blocking::Client;
use runtime::{
edit_file, execute_bash, glob_search, grep_search, load_system_prompt, read_file, write_file,
ApiClient, ApiRequest, AssistantEvent, BashCommandInput, ConfigLoader, ContentBlock,
ConversationMessage, ConversationRuntime, GrepSearchInput, MessageRole, PermissionMode,
PermissionPolicy, RuntimeError, Session, TokenUsage, ToolError, ToolExecutor,
edit_file, execute_bash, glob_search, grep_search, read_file, write_file, BashCommandInput,
GrepSearchInput, PermissionMode,
};
use serde::{Deserialize, Serialize};
use serde_json::{json, Value};
@@ -69,7 +62,11 @@ pub fn mvp_tool_specs() -> Vec<ToolSpec> {
"timeout": { "type": "integer", "minimum": 1 },
"description": { "type": "string" },
"run_in_background": { "type": "boolean" },
"dangerouslyDisableSandbox": { "type": "boolean" }
"dangerouslyDisableSandbox": { "type": "boolean" },
"namespaceRestrictions": { "type": "boolean" },
"isolateNetwork": { "type": "boolean" },
"filesystemMode": { "type": "string", "enum": ["off", "workspace-only", "allow-list"] },
"allowedMounts": { "type": "array", "items": { "type": "string" } }
},
"required": ["command"],
"additionalProperties": false
@@ -241,8 +238,7 @@ pub fn mvp_tool_specs() -> Vec<ToolSpec> {
},
ToolSpec {
name: "Agent",
description:
"Launch and execute a specialized child agent conversation with bounded recursion.",
description: "Launch a specialized agent task and persist its handoff metadata.",
input_schema: json!({
"type": "object",
"properties": {
@@ -250,8 +246,7 @@ pub fn mvp_tool_specs() -> Vec<ToolSpec> {
"prompt": { "type": "string" },
"subagent_type": { "type": "string" },
"name": { "type": "string" },
"model": { "type": "string" },
"max_depth": { "type": "integer", "minimum": 0 }
"model": { "type": "string" }
},
"required": ["description", "prompt"],
"additionalProperties": false
@@ -588,7 +583,6 @@ struct AgentInput {
subagent_type: Option<String>,
name: Option<String>,
model: Option<String>,
max_depth: Option<usize>,
}
#[derive(Debug, Deserialize)]
@@ -722,16 +716,6 @@ struct AgentOutput {
subagent_type: Option<String>,
model: Option<String>,
status: String,
#[serde(rename = "maxDepth")]
max_depth: usize,
#[serde(rename = "depth")]
depth: usize,
#[serde(rename = "result")]
result: Option<String>,
#[serde(rename = "assistantMessages")]
assistant_messages: Vec<String>,
#[serde(rename = "toolResults")]
tool_results: Vec<AgentToolResult>,
#[serde(rename = "outputFile")]
output_file: String,
#[serde(rename = "manifestFile")]
@@ -740,15 +724,6 @@ struct AgentOutput {
created_at: String,
}
#[derive(Debug, Clone, Serialize, Deserialize)]
struct AgentToolResult {
#[serde(rename = "toolName")]
tool_name: String,
output: String,
#[serde(rename = "isError")]
is_error: bool,
}
#[derive(Debug, Serialize)]
struct ToolSearchOutput {
matches: Vec<String>,
@@ -1360,14 +1335,6 @@ fn execute_agent(input: AgentInput) -> Result<AgentOutput, String> {
return Err(String::from("prompt must not be empty"));
}
let depth = current_agent_depth()?;
let max_depth = input.max_depth.unwrap_or(3);
if depth >= max_depth {
return Err(format!(
"Agent max_depth exceeded: current depth {depth} reached limit {max_depth}"
));
}
let agent_id = make_agent_id();
let output_dir = agent_store_dir()?;
std::fs::create_dir_all(&output_dir).map_err(|error| error.to_string())?;
@@ -1381,31 +1348,35 @@ fn execute_agent(input: AgentInput) -> Result<AgentOutput, String> {
.filter(|name| !name.is_empty())
.unwrap_or_else(|| slugify_agent_name(&input.description));
let created_at = iso8601_now();
let model = input.model.clone().or_else(agent_default_model);
let child_result = with_agent_depth(depth + 1, || {
run_child_agent_conversation(&input.prompt, model.clone(), max_depth)
})?;
let output_contents = format!(
"# Agent Task
- id: {}
- name: {}
- description: {}
- subagent_type: {}
- created_at: {}
## Prompt
{}
",
agent_id, agent_name, input.description, normalized_subagent_type, created_at, input.prompt
);
std::fs::write(&output_file, output_contents).map_err(|error| error.to_string())?;
let manifest = AgentOutput {
agent_id,
name: agent_name,
description: input.description,
subagent_type: Some(normalized_subagent_type),
model,
status: String::from("completed"),
max_depth,
depth,
result: child_result.result.clone(),
assistant_messages: child_result.assistant_messages.clone(),
tool_results: child_result.tool_results.clone(),
model: input.model,
status: String::from("queued"),
output_file: output_file.display().to_string(),
manifest_file: manifest_file.display().to_string(),
created_at,
};
let output_contents = render_agent_output(&manifest);
std::fs::write(&output_file, output_contents).map_err(|error| error.to_string())?;
std::fs::write(
&manifest_file,
serde_json::to_string_pretty(&manifest).map_err(|error| error.to_string())?,
@@ -1415,461 +1386,6 @@ fn execute_agent(input: AgentInput) -> Result<AgentOutput, String> {
Ok(manifest)
}
#[derive(Debug, Clone)]
struct ChildConversationResult {
result: Option<String>,
assistant_messages: Vec<String>,
tool_results: Vec<AgentToolResult>,
}
fn run_child_agent_conversation(
prompt: &str,
model: Option<String>,
_max_depth: usize,
) -> Result<ChildConversationResult, String> {
let mut runtime = ConversationRuntime::new(
Session::new(),
build_agent_api_client(model.unwrap_or_else(default_agent_model))?,
AgentToolExecutor,
agent_permission_policy(),
build_agent_system_prompt()?,
)
.with_max_iterations(16);
let summary = runtime
.run_turn(prompt, None)
.map_err(|error| error.to_string())?;
let assistant_messages = summary
.assistant_messages
.iter()
.filter_map(extract_message_text)
.collect::<Vec<_>>();
let tool_results = summary
.tool_results
.iter()
.filter_map(extract_agent_tool_result)
.collect::<Vec<_>>();
let result = assistant_messages.last().cloned();
Ok(ChildConversationResult {
result,
assistant_messages,
tool_results,
})
}
fn render_agent_output(output: &AgentOutput) -> String {
let mut lines = vec![
"# Agent Task".to_string(),
String::new(),
format!("- id: {}", output.agent_id),
format!("- name: {}", output.name),
format!("- description: {}", output.description),
format!(
"- subagent_type: {}",
output.subagent_type.as_deref().unwrap_or("general-purpose")
),
format!("- status: {}", output.status),
format!("- depth: {}", output.depth),
format!("- max_depth: {}", output.max_depth),
format!("- created_at: {}", output.created_at),
String::new(),
"## Result".to_string(),
String::new(),
output
.result
.clone()
.unwrap_or_else(|| String::from("<no final assistant text>")),
];
if !output.tool_results.is_empty() {
lines.push(String::new());
lines.push("## Tool Results".to_string());
lines.push(String::new());
lines.extend(output.tool_results.iter().map(|result| {
format!(
"- {} [{}]: {}",
result.tool_name,
if result.is_error { "error" } else { "ok" },
result.output
)
}));
}
lines.join("\n")
}
fn current_agent_depth() -> Result<usize, String> {
std::env::var("CLAWD_AGENT_DEPTH")
.ok()
.map(|value| {
value
.parse::<usize>()
.map_err(|error| format!("invalid CLAWD_AGENT_DEPTH: {error}"))
})
.transpose()
.map(|value| value.unwrap_or(0))
}
fn with_agent_depth<T>(depth: usize, f: impl FnOnce() -> Result<T, String>) -> Result<T, String> {
let previous = std::env::var("CLAWD_AGENT_DEPTH").ok();
std::env::set_var("CLAWD_AGENT_DEPTH", depth.to_string());
let result = f();
if let Some(previous) = previous {
std::env::set_var("CLAWD_AGENT_DEPTH", previous);
} else {
std::env::remove_var("CLAWD_AGENT_DEPTH");
}
result
}
fn agent_default_model() -> Option<String> {
std::env::var("CLAWD_MODEL")
.ok()
.filter(|value| !value.trim().is_empty())
}
fn default_agent_model() -> String {
agent_default_model().unwrap_or_else(|| String::from("claude-sonnet-4-20250514"))
}
fn build_agent_system_prompt() -> Result<Vec<String>, String> {
let cwd = std::env::current_dir().map_err(|error| error.to_string())?;
let date = std::env::var("CLAWD_CURRENT_DATE").unwrap_or_else(|_| String::from("2026-04-01"));
load_system_prompt(cwd, &date, std::env::consts::OS, "unknown")
.map_err(|error| error.to_string())
}
fn agent_permission_policy() -> PermissionPolicy {
mvp_tool_specs().into_iter().fold(
PermissionPolicy::new(PermissionMode::DangerFullAccess),
|policy, spec| policy.with_tool_requirement(spec.name, spec.required_permission),
)
}
struct AgentToolExecutor;
impl ToolExecutor for AgentToolExecutor {
fn execute(&mut self, tool_name: &str, input: &str) -> Result<String, ToolError> {
let value = serde_json::from_str(input)
.map_err(|error| ToolError::new(format!("invalid tool input JSON: {error}")))?;
execute_tool(tool_name, &value).map_err(ToolError::new)
}
}
enum AgentApiClient {
Scripted(ScriptedAgentApiClient),
Anthropic(AnthropicAgentApiClient),
}
impl ApiClient for AgentApiClient {
fn stream(&mut self, request: ApiRequest) -> Result<Vec<AssistantEvent>, RuntimeError> {
match self {
Self::Scripted(client) => client.stream(request),
Self::Anthropic(client) => client.stream(request),
}
}
}
fn build_agent_api_client(model: String) -> Result<AgentApiClient, String> {
if let Some(script) = std::env::var("CLAWD_AGENT_TEST_SCRIPT")
.ok()
.filter(|value| !value.trim().is_empty())
{
return Ok(AgentApiClient::Scripted(ScriptedAgentApiClient::new(
&script,
)?));
}
Ok(AgentApiClient::Anthropic(AnthropicAgentApiClient::new(
model,
)?))
}
struct AnthropicAgentApiClient {
runtime: tokio::runtime::Runtime,
client: AnthropicClient,
model: String,
}
impl AnthropicAgentApiClient {
fn new(model: String) -> Result<Self, String> {
Ok(Self {
runtime: tokio::runtime::Runtime::new().map_err(|error| error.to_string())?,
client: AnthropicClient::from_auth(resolve_agent_auth_source()?),
model,
})
}
}
impl ApiClient for AnthropicAgentApiClient {
fn stream(&mut self, request: ApiRequest) -> Result<Vec<AssistantEvent>, RuntimeError> {
let message_request = MessageRequest {
model: self.model.clone(),
max_tokens: 32,
messages: convert_agent_messages(&request.messages),
system: (!request.system_prompt.is_empty()).then(|| {
request.system_prompt.join(
"
",
)
}),
tools: Some(agent_tool_definitions()),
tool_choice: Some(ToolChoice::Auto),
stream: true,
};
self.runtime.block_on(async {
let mut stream = self
.client
.stream_message(&message_request)
.await
.map_err(|error| RuntimeError::new(error.to_string()))?;
let mut events = Vec::new();
let mut pending_tool: Option<(String, String, String)> = None;
let mut saw_stop = false;
while let Some(event) = stream
.next_event()
.await
.map_err(|error| RuntimeError::new(error.to_string()))?
{
match event {
ApiStreamEvent::MessageStart(start) => {
push_agent_output_blocks(
start.message.content,
&mut events,
&mut pending_tool,
);
}
ApiStreamEvent::ContentBlockStart(start) => {
push_agent_output_block(
start.content_block,
&mut events,
&mut pending_tool,
);
}
ApiStreamEvent::ContentBlockDelta(delta) => match delta.delta {
ContentBlockDelta::TextDelta { text } => {
if !text.is_empty() {
events.push(AssistantEvent::TextDelta(text));
}
}
ContentBlockDelta::InputJsonDelta { partial_json } => {
if let Some((_, _, input)) = &mut pending_tool {
input.push_str(&partial_json);
}
}
},
ApiStreamEvent::ContentBlockStop(_) => {
if let Some((id, name, input)) = pending_tool.take() {
events.push(AssistantEvent::ToolUse { id, name, input });
}
}
ApiStreamEvent::MessageDelta(delta) => {
events.push(AssistantEvent::Usage(TokenUsage {
input_tokens: delta.usage.input_tokens,
output_tokens: delta.usage.output_tokens,
cache_creation_input_tokens: delta.usage.cache_creation_input_tokens,
cache_read_input_tokens: delta.usage.cache_read_input_tokens,
}));
}
ApiStreamEvent::MessageStop(_) => {
saw_stop = true;
events.push(AssistantEvent::MessageStop);
}
}
}
if !saw_stop {
events.push(AssistantEvent::MessageStop);
}
Ok(events)
})
}
}
fn resolve_agent_auth_source() -> Result<api::AuthSource, String> {
resolve_startup_auth_source(|| {
let cwd = std::env::current_dir().map_err(api::ApiError::from)?;
let config = ConfigLoader::default_for(&cwd).load().map_err(|error| {
api::ApiError::Auth(format!("failed to load runtime OAuth config: {error}"))
})?;
Ok(config.oauth().cloned())
})
.map_err(|error| error.to_string())
}
fn agent_tool_definitions() -> Vec<ToolDefinition> {
mvp_tool_specs()
.into_iter()
.map(|spec| ToolDefinition {
name: spec.name.to_string(),
description: Some(spec.description.to_string()),
input_schema: spec.input_schema,
})
.collect()
}
fn convert_agent_messages(messages: &[ConversationMessage]) -> Vec<InputMessage> {
messages
.iter()
.filter_map(|message| {
let role = match message.role {
MessageRole::System | MessageRole::User | MessageRole::Tool => "user",
MessageRole::Assistant => "assistant",
};
let content = message
.blocks
.iter()
.map(|block| match block {
ContentBlock::Text { text } => InputContentBlock::Text { text: text.clone() },
ContentBlock::ToolUse { id, name, input } => InputContentBlock::ToolUse {
id: id.clone(),
name: name.clone(),
input: serde_json::from_str(input)
.unwrap_or_else(|_| serde_json::json!({ "raw": input })),
},
ContentBlock::ToolResult {
tool_use_id,
output,
is_error,
..
} => InputContentBlock::ToolResult {
tool_use_id: tool_use_id.clone(),
content: vec![ToolResultContentBlock::Text {
text: output.clone(),
}],
is_error: *is_error,
},
})
.collect::<Vec<_>>();
(!content.is_empty()).then(|| InputMessage {
role: role.to_string(),
content,
})
})
.collect()
}
fn push_agent_output_blocks(
blocks: Vec<OutputContentBlock>,
events: &mut Vec<AssistantEvent>,
pending_tool: &mut Option<(String, String, String)>,
) {
for block in blocks {
push_agent_output_block(block, events, pending_tool);
if let Some((id, name, input)) = pending_tool.take() {
events.push(AssistantEvent::ToolUse { id, name, input });
}
}
}
fn push_agent_output_block(
block: OutputContentBlock,
events: &mut Vec<AssistantEvent>,
pending_tool: &mut Option<(String, String, String)>,
) {
match block {
OutputContentBlock::Text { text } => {
if !text.is_empty() {
events.push(AssistantEvent::TextDelta(text));
}
}
OutputContentBlock::ToolUse { id, name, input } => {
*pending_tool = Some((id, name, input.to_string()));
}
}
}
#[derive(Debug)]
struct ScriptedAgentApiClient {
turns: Vec<Vec<ScriptedAgentEvent>>,
call_count: usize,
}
impl ScriptedAgentApiClient {
fn new(script: &str) -> Result<Self, String> {
let turns = serde_json::from_str(script).map_err(|error| error.to_string())?;
Ok(Self {
turns,
call_count: 0,
})
}
}
impl ApiClient for ScriptedAgentApiClient {
fn stream(&mut self, _request: ApiRequest) -> Result<Vec<AssistantEvent>, RuntimeError> {
if self.call_count >= self.turns.len() {
return Err(RuntimeError::new("scripted agent client exhausted"));
}
let events = self.turns[self.call_count]
.iter()
.map(ScriptedAgentEvent::to_runtime_event)
.chain(std::iter::once(AssistantEvent::MessageStop))
.collect();
self.call_count += 1;
Ok(events)
}
}
#[derive(Debug, Clone, Deserialize)]
#[serde(tag = "type", rename_all = "snake_case")]
enum ScriptedAgentEvent {
Text {
text: String,
},
ToolUse {
id: String,
name: String,
input: Value,
},
}
impl ScriptedAgentEvent {
fn to_runtime_event(&self) -> AssistantEvent {
match self {
Self::Text { text } => AssistantEvent::TextDelta(text.clone()),
Self::ToolUse { id, name, input } => AssistantEvent::ToolUse {
id: id.clone(),
name: name.clone(),
input: input.to_string(),
},
}
}
}
fn extract_message_text(message: &ConversationMessage) -> Option<String> {
let text = message
.blocks
.iter()
.filter_map(|block| match block {
ContentBlock::Text { text } => Some(text.as_str()),
_ => None,
})
.collect::<String>();
(!text.is_empty()).then_some(text)
}
fn extract_agent_tool_result(message: &ConversationMessage) -> Option<AgentToolResult> {
message.blocks.iter().find_map(|block| match block {
ContentBlock::ToolResult {
tool_name,
output,
is_error,
..
} => Some(AgentToolResult {
tool_name: tool_name.clone(),
output: output.clone(),
is_error: *is_error,
}),
_ => None,
})
}
#[allow(clippy::needless_pass_by_value)]
fn execute_tool_search(input: ToolSearchInput) -> ToolSearchOutput {
let deferred = deferred_tool_specs();
@@ -2702,6 +2218,7 @@ fn execute_shell_command(
structured_content: None,
persisted_output_path: None,
persisted_output_size: None,
sandbox_status: None,
});
}
@@ -2739,6 +2256,7 @@ fn execute_shell_command(
structured_content: None,
persisted_output_path: None,
persisted_output_size: None,
sandbox_status: None,
});
}
if started.elapsed() >= Duration::from_millis(timeout_ms) {
@@ -2769,6 +2287,7 @@ Command exceeded timeout of {timeout_ms} ms",
structured_content: None,
persisted_output_path: None,
persisted_output_size: None,
sandbox_status: None,
});
}
std::thread::sleep(Duration::from_millis(10));
@@ -2795,6 +2314,7 @@ Command exceeded timeout of {timeout_ms} ms",
structured_content: None,
persisted_output_path: None,
persisted_output_size: None,
sandbox_status: None,
})
}
@@ -3251,28 +2771,12 @@ mod tests {
}
#[test]
fn agent_executes_child_conversation_and_persists_results() {
fn agent_persists_handoff_metadata() {
let _guard = env_lock()
.lock()
.unwrap_or_else(std::sync::PoisonError::into_inner);
let dir = temp_path("agent-store");
std::env::set_var("CLAWD_AGENT_STORE", &dir);
std::env::set_var(
"CLAWD_AGENT_TEST_SCRIPT",
serde_json::to_string(&vec![
vec![json!({
"type": "tool_use",
"id": "tool-1",
"name": "StructuredOutput",
"input": {"ok": true, "items": [1, 2, 3]}
})],
vec![json!({
"type": "text",
"text": "Child agent completed successfully."
})],
])
.expect("script json"),
);
let result = execute_tool(
"Agent",
@@ -3284,35 +2788,22 @@ mod tests {
}),
)
.expect("Agent should succeed");
std::env::remove_var("CLAWD_AGENT_TEST_SCRIPT");
std::env::remove_var("CLAWD_AGENT_STORE");
let output: serde_json::Value = serde_json::from_str(&result).expect("valid json");
assert_eq!(output["name"], "ship-audit");
assert_eq!(output["subagentType"], "Explore");
assert_eq!(output["status"], "completed");
assert_eq!(output["depth"], 0);
assert_eq!(output["maxDepth"], 3);
assert_eq!(output["result"], "Child agent completed successfully.");
assert_eq!(output["toolResults"][0]["toolName"], "StructuredOutput");
assert_eq!(output["toolResults"][0]["isError"], false);
assert_eq!(output["status"], "queued");
assert!(output["createdAt"].as_str().is_some());
let manifest_file = output["manifestFile"].as_str().expect("manifest file");
let output_file = output["outputFile"].as_str().expect("output file");
let contents = std::fs::read_to_string(output_file).expect("agent file exists");
let manifest_contents =
std::fs::read_to_string(manifest_file).expect("manifest file exists");
assert!(contents.contains("Child agent completed successfully."));
assert!(contents.contains("StructuredOutput [ok]"));
assert!(contents.contains("Audit the branch"));
assert!(contents.contains("Check tests and outstanding work."));
assert!(manifest_contents.contains("\"subagentType\": \"Explore\""));
std::env::set_var(
"CLAWD_AGENT_TEST_SCRIPT",
serde_json::to_string(&vec![vec![json!({
"type": "text",
"text": "Normalized alias check."
})]])
.expect("script json"),
);
let normalized = execute_tool(
"Agent",
&json!({
@@ -3322,19 +2813,10 @@ mod tests {
}),
)
.expect("Agent should normalize built-in aliases");
std::env::remove_var("CLAWD_AGENT_TEST_SCRIPT");
let normalized_output: serde_json::Value =
serde_json::from_str(&normalized).expect("valid json");
assert_eq!(normalized_output["subagentType"], "Explore");
std::env::set_var(
"CLAWD_AGENT_TEST_SCRIPT",
serde_json::to_string(&vec![vec![json!({
"type": "text",
"text": "Name normalization check."
})]])
.expect("script json"),
);
let named = execute_tool(
"Agent",
&json!({
@@ -3344,14 +2826,13 @@ mod tests {
}),
)
.expect("Agent should normalize explicit names");
std::env::remove_var("CLAWD_AGENT_TEST_SCRIPT");
let named_output: serde_json::Value = serde_json::from_str(&named).expect("valid json");
assert_eq!(named_output["name"], "ship-audit");
let _ = std::fs::remove_dir_all(dir);
}
#[test]
fn agent_rejects_blank_required_fields_and_enforces_max_depth() {
fn agent_rejects_blank_required_fields() {
let missing_description = execute_tool(
"Agent",
&json!({
@@ -3371,22 +2852,6 @@ mod tests {
)
.expect_err("blank prompt should fail");
assert!(missing_prompt.contains("prompt must not be empty"));
let _guard = env_lock()
.lock()
.unwrap_or_else(std::sync::PoisonError::into_inner);
std::env::set_var("CLAWD_AGENT_DEPTH", "1");
let depth_error = execute_tool(
"Agent",
&json!({
"description": "Nested agent",
"prompt": "Do nested work.",
"max_depth": 1
}),
)
.expect_err("max depth should fail");
std::env::remove_var("CLAWD_AGENT_DEPTH");
assert!(depth_error.contains("max_depth exceeded"));
}
#[test]