summaryrefslogtreecommitdiff
path: root/makima/src/server/handlers
diff options
context:
space:
mode:
Diffstat (limited to 'makima/src/server/handlers')
-rw-r--r--makima/src/server/handlers/contract_chat.rs6
-rw-r--r--makima/src/server/handlers/contracts.rs1
-rw-r--r--makima/src/server/handlers/mesh.rs4
-rw-r--r--makima/src/server/handlers/mesh_chat.rs1
-rw-r--r--makima/src/server/handlers/mesh_red_team.rs497
-rw-r--r--makima/src/server/handlers/mesh_supervisor.rs292
-rw-r--r--makima/src/server/handlers/mod.rs1
-rw-r--r--makima/src/server/handlers/transcript_analysis.rs4
8 files changed, 806 insertions, 0 deletions
diff --git a/makima/src/server/handlers/contract_chat.rs b/makima/src/server/handlers/contract_chat.rs
index e6ee8d4..dac806a 100644
--- a/makima/src/server/handlers/contract_chat.rs
+++ b/makima/src/server/handlers/contract_chat.rs
@@ -1362,6 +1362,7 @@ async fn handle_contract_request(
continue_from_task_id: None,
copy_files: None,
is_supervisor: false,
+ is_red_team: false,
checkpoint_sha: None,
branched_from_task_id: None,
conversation_history: None,
@@ -1458,6 +1459,7 @@ async fn handle_contract_request(
continue_from_task_id: None,
copy_files: None,
is_supervisor: false,
+ is_red_team: false,
checkpoint_sha: None,
branched_from_task_id: None,
conversation_history: None,
@@ -2190,6 +2192,7 @@ async fn handle_contract_request(
continue_from_task_id: previous_task_id,
copy_files: None,
is_supervisor: false,
+ is_red_team: false,
checkpoint_sha: None,
branched_from_task_id: None,
conversation_history: None,
@@ -2586,6 +2589,8 @@ async fn handle_contract_request(
autonomous_loop: None,
phase_guard: None,
local_only: None,
+ red_team_enabled: None,
+ red_team_prompt: None,
};
let contract = match repository::create_contract_for_owner(pool, owner_id, contract_req).await {
@@ -2707,6 +2712,7 @@ async fn handle_contract_request(
continue_from_task_id: None,
copy_files: None,
is_supervisor: false,
+ is_red_team: false,
checkpoint_sha: None,
branched_from_task_id: None,
conversation_history: None,
diff --git a/makima/src/server/handlers/contracts.rs b/makima/src/server/handlers/contracts.rs
index 3498063..3ad38da 100644
--- a/makima/src/server/handlers/contracts.rs
+++ b/makima/src/server/handlers/contracts.rs
@@ -295,6 +295,7 @@ pub async fn create_contract(
continue_from_task_id: None,
copy_files: None,
is_supervisor: true,
+ is_red_team: false,
checkpoint_sha: None,
priority: 0,
merge_mode: None,
diff --git a/makima/src/server/handlers/mesh.rs b/makima/src/server/handlers/mesh.rs
index 19958e7..c4d862c 100644
--- a/makima/src/server/handlers/mesh.rs
+++ b/makima/src/server/handlers/mesh.rs
@@ -2239,6 +2239,7 @@ pub async fn reassign_task(
plan: updated_plan.clone(),
parent_task_id: task.parent_task_id,
is_supervisor: task.is_supervisor,
+ is_red_team: task.is_red_team,
priority: task.priority,
repository_url: task.repository_url.clone(),
base_branch: task.base_branch.clone(),
@@ -3010,6 +3011,7 @@ pub async fn fork_task(
plan: req.new_task_plan.clone(),
parent_task_id: None, // Forked tasks are independent
is_supervisor: false,
+ is_red_team: false,
priority: task.priority,
repository_url: task.repository_url.clone(),
base_branch: task.base_branch.clone(),
@@ -3167,6 +3169,7 @@ pub async fn resume_from_checkpoint(
plan: req.plan,
parent_task_id: None,
is_supervisor: false,
+ is_red_team: false,
priority: task.priority,
repository_url: task.repository_url.clone(),
base_branch: task.base_branch.clone(),
@@ -3502,6 +3505,7 @@ pub async fn branch_task(
plan: req.message,
parent_task_id: None,
is_supervisor: false,
+ is_red_team: false,
priority: source_task.priority,
repository_url: source_task.repository_url.clone(),
base_branch: source_task.base_branch.clone(),
diff --git a/makima/src/server/handlers/mesh_chat.rs b/makima/src/server/handlers/mesh_chat.rs
index eb35728..ed6cfc0 100644
--- a/makima/src/server/handlers/mesh_chat.rs
+++ b/makima/src/server/handlers/mesh_chat.rs
@@ -1017,6 +1017,7 @@ async fn handle_mesh_request(
continue_from_task_id: None,
copy_files: None,
is_supervisor: false,
+ is_red_team: false,
checkpoint_sha: None,
branched_from_task_id: None,
conversation_history: None,
diff --git a/makima/src/server/handlers/mesh_red_team.rs b/makima/src/server/handlers/mesh_red_team.rs
new file mode 100644
index 0000000..c5af60e
--- /dev/null
+++ b/makima/src/server/handlers/mesh_red_team.rs
@@ -0,0 +1,497 @@
+//! HTTP handlers for red team mesh operations.
+//!
+//! These endpoints are used by red team tasks (via the makima CLI) to notify
+//! supervisors of potential issues and query their own status.
+
+use axum::{
+ extract::State,
+ http::{HeaderMap, StatusCode},
+ response::IntoResponse,
+ Json,
+};
+use serde::{Deserialize, Serialize};
+use utoipa::ToSchema;
+use uuid::Uuid;
+
+use crate::db::repository;
+use crate::server::handlers::mesh::{extract_auth, AuthSource};
+use crate::server::messages::ApiError;
+use crate::server::state::{DaemonCommand, SharedState};
+
+// =============================================================================
+// Request/Response Types
+// =============================================================================
+
+/// Severity level for red team notifications.
+#[derive(Debug, Clone, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "lowercase")]
+pub enum RedTeamSeverity {
+ /// Informational notice - minor issue or suggestion
+ Info,
+ /// Warning - potential problem that should be reviewed
+ Warning,
+ /// Critical - serious issue requiring immediate attention
+ Critical,
+}
+
+impl Default for RedTeamSeverity {
+ fn default() -> Self {
+ Self::Warning
+ }
+}
+
+impl std::fmt::Display for RedTeamSeverity {
+ fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+ match self {
+ Self::Info => write!(f, "INFO"),
+ Self::Warning => write!(f, "WARNING"),
+ Self::Critical => write!(f, "CRITICAL"),
+ }
+ }
+}
+
+/// Request to notify the supervisor of a potential issue.
+#[derive(Debug, Deserialize, ToSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct RedTeamNotifyRequest {
+ /// The issue description/message to send to the supervisor
+ pub message: String,
+ /// Severity level of the issue
+ #[serde(default)]
+ pub severity: RedTeamSeverity,
+ /// ID of the task being reviewed (optional - if not provided, assumes general contract concern)
+ pub related_task_id: Option<Uuid>,
+ /// File path related to the issue (optional)
+ pub file_path: Option<String>,
+ /// Additional context about the issue
+ pub context: Option<String>,
+}
+
+/// Response from the notify endpoint.
+#[derive(Debug, Serialize, ToSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct RedTeamNotifyResponse {
+ /// Unique ID for this notification
+ pub notification_id: Uuid,
+ /// Whether the notification was successfully delivered to the supervisor
+ pub delivered: bool,
+ /// The supervisor task ID that received the notification
+ pub supervisor_task_id: Option<Uuid>,
+}
+
+/// Response from the status endpoint.
+#[derive(Debug, Serialize, ToSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct RedTeamStatusResponse {
+ /// Contract ID being monitored
+ pub contract_id: Uuid,
+ /// Red team task ID
+ pub red_team_task_id: Uuid,
+ /// Current task status
+ pub status: String,
+ /// Number of notifications sent so far
+ pub notifications_sent: i64,
+}
+
+/// Red team notification record stored in database.
+#[derive(Debug, Clone, Serialize, ToSchema)]
+#[serde(rename_all = "camelCase")]
+pub struct RedTeamNotification {
+ pub id: Uuid,
+ pub red_team_task_id: Uuid,
+ pub contract_id: Uuid,
+ pub message: String,
+ pub severity: String,
+ pub related_task_id: Option<Uuid>,
+ pub file_path: Option<String>,
+ pub context: Option<String>,
+ pub delivered: bool,
+ pub created_at: chrono::DateTime<chrono::Utc>,
+}
+
+// =============================================================================
+// Helper Functions
+// =============================================================================
+
+/// Verify the request comes from a red team task and extract ownership info.
+///
+/// Returns (task_id, owner_id, contract_id) on success.
+async fn verify_red_team_auth(
+ state: &SharedState,
+ headers: &HeaderMap,
+) -> Result<(Uuid, Uuid, Uuid), (StatusCode, Json<ApiError>)> {
+ let auth = extract_auth(state, headers);
+
+ let task_id = match auth {
+ AuthSource::ToolKey(task_id) => task_id,
+ _ => {
+ return Err((
+ StatusCode::UNAUTHORIZED,
+ Json(ApiError::new(
+ "UNAUTHORIZED",
+ "Red team endpoints require tool key auth",
+ )),
+ ));
+ }
+ };
+
+ // Get the task to verify it's a red team task and get owner_id
+ let pool = state.db_pool.as_ref().ok_or_else(|| {
+ (
+ StatusCode::SERVICE_UNAVAILABLE,
+ Json(ApiError::new("DB_UNAVAILABLE", "Database not configured")),
+ )
+ })?;
+
+ let task = repository::get_task(pool, task_id)
+ .await
+ .map_err(|e| {
+ tracing::error!(error = %e, "Failed to get red team task");
+ (
+ StatusCode::INTERNAL_SERVER_ERROR,
+ Json(ApiError::new("DB_ERROR", "Failed to verify red team task")),
+ )
+ })?
+ .ok_or_else(|| {
+ (
+ StatusCode::NOT_FOUND,
+ Json(ApiError::new("NOT_FOUND", "Task not found")),
+ )
+ })?;
+
+ // Verify task is a red team task
+ // NOTE: This requires the is_red_team field to be added to the Task struct.
+ // For now, we check if the task name contains "red-team" or "red_team" as a fallback.
+ let is_red_team = task.name.to_lowercase().contains("red-team")
+ || task.name.to_lowercase().contains("red_team")
+ || task.name.to_lowercase().contains("redteam");
+
+ if !is_red_team {
+ return Err((
+ StatusCode::FORBIDDEN,
+ Json(ApiError::new(
+ "NOT_RED_TEAM",
+ "Only red team tasks can use these endpoints",
+ )),
+ ));
+ }
+
+ // Red team tasks must be associated with a contract
+ let contract_id = task.contract_id.ok_or_else(|| {
+ (
+ StatusCode::BAD_REQUEST,
+ Json(ApiError::new(
+ "NO_CONTRACT",
+ "Red team task must be associated with a contract",
+ )),
+ )
+ })?;
+
+ Ok((task_id, task.owner_id, contract_id))
+}
+
+/// Format an alert message for the supervisor.
+///
+/// Creates a formatted alert with clear visual markers to grab attention.
+fn format_alert_message(
+ severity: &RedTeamSeverity,
+ message: &str,
+ related_task_id: Option<Uuid>,
+ file_path: Option<&str>,
+ context: Option<&str>,
+) -> String {
+ let severity_marker = match severity {
+ RedTeamSeverity::Info => "â„šī¸",
+ RedTeamSeverity::Warning => "âš ī¸",
+ RedTeamSeverity::Critical => "🚨",
+ };
+
+ let border = match severity {
+ RedTeamSeverity::Info => "─".repeat(60),
+ RedTeamSeverity::Warning => "━".repeat(60),
+ RedTeamSeverity::Critical => "═".repeat(60),
+ };
+
+ let mut alert = format!(
+ r#"
+{}
+{} [RED TEAM ALERT] - {}
+{}
+
+Issue: {}
+"#,
+ border, severity_marker, severity, border, message
+ );
+
+ if let Some(task_id) = related_task_id {
+ alert.push_str(&format!("\nRelated Task: {}\n", task_id));
+ }
+
+ if let Some(path) = file_path {
+ alert.push_str(&format!("File: {}\n", path));
+ }
+
+ if let Some(ctx) = context {
+ alert.push_str(&format!("\nContext:\n{}\n", ctx));
+ }
+
+ // Add action suggestions based on severity
+ let actions = match severity {
+ RedTeamSeverity::Info => {
+ "Suggested Actions:\n- Review when convenient\n- Consider if changes are needed"
+ }
+ RedTeamSeverity::Warning => {
+ "Suggested Actions:\n- Review the flagged item soon\n- Check if this deviates from the contract\n- Consider pausing related work until reviewed"
+ }
+ RedTeamSeverity::Critical => {
+ "Suggested Actions:\n- STOP related work immediately\n- Review the flagged item urgently\n- Verify compliance with contract requirements\n- Consider reverting recent changes if necessary"
+ }
+ };
+
+ alert.push_str(&format!("\n{}\n{}\n", actions, border));
+
+ alert
+}
+
+// =============================================================================
+// Handlers
+// =============================================================================
+
+/// Notify the supervisor of a potential issue.
+///
+/// POST /api/v1/mesh/red-team/notify
+///
+/// This endpoint allows red team tasks to alert supervisors about issues they've
+/// identified during code review. The notification is sent as a message to the
+/// supervisor task.
+#[utoipa::path(
+ post,
+ path = "/api/v1/mesh/red-team/notify",
+ request_body = RedTeamNotifyRequest,
+ responses(
+ (status = 200, description = "Notification sent", body = RedTeamNotifyResponse),
+ (status = 401, description = "Unauthorized - tool key required"),
+ (status = 403, description = "Forbidden - not a red team task"),
+ (status = 404, description = "Task not found"),
+ (status = 503, description = "Database not available"),
+ (status = 500, description = "Internal server error"),
+ ),
+ security(
+ ("tool_key" = [])
+ ),
+ tag = "Mesh Red Team"
+)]
+pub async fn notify_supervisor(
+ State(state): State<SharedState>,
+ headers: HeaderMap,
+ Json(request): Json<RedTeamNotifyRequest>,
+) -> impl IntoResponse {
+ let (red_team_task_id, owner_id, contract_id) =
+ match verify_red_team_auth(&state, &headers).await {
+ Ok(ids) => ids,
+ Err(e) => return e.into_response(),
+ };
+
+ let pool = state.db_pool.as_ref().unwrap();
+
+ // Generate notification ID
+ let notification_id = Uuid::new_v4();
+
+ // Get the contract to find the supervisor task
+ let contract = match repository::get_contract_for_owner(pool, contract_id, owner_id).await {
+ Ok(Some(c)) => c,
+ Ok(None) => {
+ return (
+ StatusCode::NOT_FOUND,
+ Json(ApiError::new("NOT_FOUND", "Contract not found")),
+ )
+ .into_response();
+ }
+ Err(e) => {
+ tracing::error!(error = %e, "Failed to get contract");
+ return (
+ StatusCode::INTERNAL_SERVER_ERROR,
+ Json(ApiError::new("DB_ERROR", "Failed to get contract")),
+ )
+ .into_response();
+ }
+ };
+
+ let supervisor_task_id = contract.supervisor_task_id;
+
+ // Format the alert message
+ let alert_message = format_alert_message(
+ &request.severity,
+ &request.message,
+ request.related_task_id,
+ request.file_path.as_deref(),
+ request.context.as_deref(),
+ );
+
+ // Record the notification in the database as a history event
+ let event_data = serde_json::json!({
+ "notification_id": notification_id.to_string(),
+ "red_team_task_id": red_team_task_id.to_string(),
+ "severity": request.severity.to_string(),
+ "message": request.message,
+ "related_task_id": request.related_task_id.map(|id| id.to_string()),
+ "file_path": request.file_path,
+ "context": request.context,
+ });
+
+ let _ = repository::record_history_event(
+ pool,
+ owner_id,
+ Some(contract_id),
+ Some(red_team_task_id),
+ "red_team_alert",
+ Some(&request.severity.to_string().to_lowercase()),
+ Some(&request.message),
+ event_data,
+ )
+ .await;
+
+ // Try to send the message to the supervisor
+ let mut delivered = false;
+ if let Some(sup_task_id) = supervisor_task_id {
+ // Get the supervisor task to find its daemon
+ if let Ok(Some(supervisor_task)) = repository::get_task(pool, sup_task_id).await {
+ if let Some(daemon_id) = supervisor_task.daemon_id {
+ // Send the alert message to the supervisor
+ let cmd = DaemonCommand::SendMessage {
+ task_id: sup_task_id,
+ message: alert_message.clone(),
+ };
+
+ if let Err(e) = state.send_daemon_command(daemon_id, cmd).await {
+ tracing::warn!(
+ error = %e,
+ supervisor_task_id = %sup_task_id,
+ daemon_id = %daemon_id,
+ "Failed to send red team alert to supervisor"
+ );
+ } else {
+ delivered = true;
+ tracing::info!(
+ notification_id = %notification_id,
+ red_team_task_id = %red_team_task_id,
+ supervisor_task_id = %sup_task_id,
+ severity = %request.severity,
+ "Red team alert delivered to supervisor"
+ );
+ }
+ } else {
+ tracing::warn!(
+ supervisor_task_id = %sup_task_id,
+ "Supervisor task has no assigned daemon - alert not delivered"
+ );
+ }
+ }
+ } else {
+ tracing::warn!(
+ contract_id = %contract_id,
+ "Contract has no supervisor task - alert not delivered"
+ );
+ }
+
+ (
+ StatusCode::OK,
+ Json(RedTeamNotifyResponse {
+ notification_id,
+ delivered,
+ supervisor_task_id,
+ }),
+ )
+ .into_response()
+}
+
+/// Get the status of the red team task.
+///
+/// GET /api/v1/mesh/red-team/status
+///
+/// Returns information about the current red team task including the contract
+/// being monitored and notification statistics.
+#[utoipa::path(
+ get,
+ path = "/api/v1/mesh/red-team/status",
+ responses(
+ (status = 200, description = "Red team status", body = RedTeamStatusResponse),
+ (status = 401, description = "Unauthorized - tool key required"),
+ (status = 403, description = "Forbidden - not a red team task"),
+ (status = 404, description = "Task not found"),
+ (status = 503, description = "Database not available"),
+ (status = 500, description = "Internal server error"),
+ ),
+ security(
+ ("tool_key" = [])
+ ),
+ tag = "Mesh Red Team"
+)]
+pub async fn get_status(
+ State(state): State<SharedState>,
+ headers: HeaderMap,
+) -> impl IntoResponse {
+ let (red_team_task_id, owner_id, contract_id) =
+ match verify_red_team_auth(&state, &headers).await {
+ Ok(ids) => ids,
+ Err(e) => return e.into_response(),
+ };
+
+ let pool = state.db_pool.as_ref().unwrap();
+
+ // Get the red team task status
+ let task = match repository::get_task(pool, red_team_task_id).await {
+ Ok(Some(t)) => t,
+ Ok(None) => {
+ return (
+ StatusCode::NOT_FOUND,
+ Json(ApiError::new("NOT_FOUND", "Red team task not found")),
+ )
+ .into_response();
+ }
+ Err(e) => {
+ tracing::error!(error = %e, "Failed to get red team task");
+ return (
+ StatusCode::INTERNAL_SERVER_ERROR,
+ Json(ApiError::new("DB_ERROR", "Failed to get task")),
+ )
+ .into_response();
+ }
+ };
+
+ // Count notifications sent by this red team task
+ // Query history_events for red_team_alert events from this task
+ let notifications_sent = match sqlx::query_scalar::<_, i64>(
+ r#"
+ SELECT COUNT(*)
+ FROM history_events
+ WHERE owner_id = $1
+ AND contract_id = $2
+ AND task_id = $3
+ AND event_type = 'red_team_alert'
+ "#,
+ )
+ .bind(owner_id)
+ .bind(contract_id)
+ .bind(red_team_task_id)
+ .fetch_one(pool)
+ .await
+ {
+ Ok(count) => count,
+ Err(e) => {
+ tracing::warn!(error = %e, "Failed to count red team notifications");
+ 0
+ }
+ };
+
+ (
+ StatusCode::OK,
+ Json(RedTeamStatusResponse {
+ contract_id,
+ red_team_task_id,
+ status: task.status,
+ notifications_sent,
+ }),
+ )
+ .into_response()
+}
diff --git a/makima/src/server/handlers/mesh_supervisor.rs b/makima/src/server/handlers/mesh_supervisor.rs
index a654a05..6d9f8fb 100644
--- a/makima/src/server/handlers/mesh_supervisor.rs
+++ b/makima/src/server/handlers/mesh_supervisor.rs
@@ -610,6 +610,7 @@ pub async fn spawn_task(
contract_id: Some(request.contract_id),
parent_task_id: request.parent_task_id,
is_supervisor: false,
+ is_red_team: false,
checkpoint_sha: request.checkpoint_sha.clone(),
merge_mode: Some("manual".to_string()),
priority: 0,
@@ -742,6 +743,67 @@ pub async fn spawn_task(
updated_fields: vec!["status".to_string(), "daemon_id".to_string()],
updated_by: "supervisor".to_string(),
});
+
+ // Check if we should spawn a red team task
+ // Conditions:
+ // 1. This is not a supervisor task
+ // 2. This is not already a red team task
+ // 3. Contract has red_team_enabled = true
+ // 4. No red team task exists for this contract yet
+ if !updated_task.is_supervisor && !updated_task.is_red_team && contract.red_team_enabled {
+ if let Some(contract_id) = updated_task.contract_id {
+ // Check if a red team task already exists
+ match repository::get_red_team_task_for_contract(pool, contract_id).await {
+ Ok(None) => {
+ // No red team task exists, spawn one
+ tracing::info!(
+ contract_id = %contract_id,
+ work_task_id = %updated_task.id,
+ "Spawning red team task for contract (first work task started)"
+ );
+ match spawn_red_team_task(
+ pool,
+ &state,
+ contract_id,
+ owner_id,
+ &contract.name,
+ &contract.phase,
+ contract.red_team_prompt.as_deref(),
+ ).await {
+ Ok(red_team_task) => {
+ tracing::info!(
+ contract_id = %contract_id,
+ red_team_task_id = %red_team_task.id,
+ "Red team task spawned successfully"
+ );
+ }
+ Err(e) => {
+ // Log error but don't fail the work task spawn
+ tracing::error!(
+ contract_id = %contract_id,
+ error = %e,
+ "Failed to spawn red team task"
+ );
+ }
+ }
+ }
+ Ok(Some(existing)) => {
+ tracing::debug!(
+ contract_id = %contract_id,
+ red_team_task_id = %existing.id,
+ "Red team task already exists for contract"
+ );
+ }
+ Err(e) => {
+ tracing::error!(
+ contract_id = %contract_id,
+ error = %e,
+ "Error checking for existing red team task"
+ );
+ }
+ }
+ }
+ }
}
break;
}
@@ -2396,3 +2458,233 @@ pub async fn rewind_conversation(
})
.into_response()
}
+
+// =============================================================================
+// Red Team Task Spawning
+// =============================================================================
+
+/// Generate the system prompt/plan for a red team task.
+///
+/// This creates detailed instructions for the red team monitor, including
+/// what to look for, severity levels, and how to report issues.
+pub fn generate_red_team_plan(
+ contract_name: &str,
+ contract_phase: &str,
+ custom_prompt: Option<&str>,
+) -> String {
+ let custom_criteria = if let Some(prompt) = custom_prompt {
+ format!(
+ r#"
+
+## Custom Review Criteria
+
+The contract owner has specified additional review criteria:
+{}
+"#,
+ prompt
+ )
+ } else {
+ String::new()
+ };
+
+ format!(
+ r#"# Red Team Monitor
+
+You are an adversarial quality reviewer for a software development contract. Your role is to monitor work task outputs in real-time and flag potential issues BEFORE they compound into larger problems.
+
+## Your Mission
+
+Monitor all task outputs and verify:
+1. **Plan Adherence**: Are tasks following the implementation plan?
+2. **Code Quality**: Does the code meet repository standards?
+3. **Contract Requirements**: Does the implementation match the specification?
+4. **Best Practices**: Are there obvious anti-patterns or issues?
+
+## Access Available
+
+You have read-only access to:
+- Task outputs (streamed in real-time)
+- Task diffs (code changes)
+- Contract specifications and plan documents
+- Repository configuration files (CONTRIBUTING.md, linting configs, etc.)
+
+## How to Monitor
+
+1. **Subscribe to task outputs**: You'll receive outputs from all work tasks
+2. **Analyze code changes**: Request diffs for completed tasks
+3. **Cross-reference**: Compare outputs against the plan and specifications
+4. **Report issues**: Use `makima red-team notify` when you detect problems
+
+## When to Notify
+
+NOTIFY the supervisor when you observe:
+- **Critical**: Security vulnerabilities, data loss risks, breaking changes
+- **High**: Significant deviations from the plan, major code quality issues
+- **Medium**: Missing tests, suboptimal implementations, minor standard violations
+- **Low**: Style inconsistencies, documentation gaps (use sparingly)
+
+## What NOT to Do
+
+- Do NOT nitpick minor style issues (that's what linters are for)
+- Do NOT block progress for trivial concerns
+- Do NOT write code or make changes yourself
+- Do NOT notify for things that are already in progress and being addressed
+- Do NOT create duplicate notifications for the same issue
+
+## Notification Format
+
+When notifying, always include:
+1. A clear, concise description of the issue
+2. The severity level (critical/high/medium/low)
+3. The related task ID if applicable
+4. The specific file or code location if known
+5. Why this matters (reference to plan, spec, or standards)
+
+## Example Notification
+
+```
+makima red-team notify "Task is implementing authentication with plaintext password storage, which contradicts the security requirements in the specification document" \
+ --severity critical \
+ --task <task_id> \
+ --file "src/auth/user.rs" \
+ --context "Specification section 3.2 requires bcrypt hashing for all passwords"
+```
+{}
+## Contract Context
+
+Contract: {}
+Phase: {}
+
+Focus your monitoring on outputs that relate to the active work tasks. Prioritize issues that could affect the success of the contract or introduce technical debt.
+"#,
+ custom_criteria, contract_name, contract_phase
+ )
+}
+
+/// Spawn a red team task for a contract.
+///
+/// This creates a red team monitor task that will observe work task outputs
+/// and can notify the supervisor about potential issues.
+pub async fn spawn_red_team_task(
+ pool: &sqlx::PgPool,
+ state: &SharedState,
+ contract_id: Uuid,
+ owner_id: Uuid,
+ contract_name: &str,
+ contract_phase: &str,
+ red_team_prompt: Option<&str>,
+) -> Result<Task, String> {
+ // Generate the red team plan/prompt
+ let plan = generate_red_team_plan(contract_name, contract_phase, red_team_prompt);
+
+ // Create task request
+ let create_req = CreateTaskRequest {
+ name: "Red Team Monitor".to_string(),
+ description: Some("Adversarial review task monitoring work task outputs".to_string()),
+ plan,
+ contract_id: Some(contract_id),
+ parent_task_id: None,
+ is_supervisor: false,
+ is_red_team: true,
+ priority: 0,
+ repository_url: None, // Red team doesn't need a repo
+ base_branch: None,
+ target_branch: None,
+ merge_mode: None,
+ target_repo_path: None,
+ completion_action: None,
+ continue_from_task_id: None,
+ copy_files: None,
+ checkpoint_sha: None,
+ branched_from_task_id: None,
+ conversation_history: None,
+ };
+
+ // Create task in DB
+ let task = repository::create_task_for_owner(pool, owner_id, create_req)
+ .await
+ .map_err(|e| format!("Failed to create red team task: {}", e))?;
+
+ tracing::info!(
+ contract_id = %contract_id,
+ red_team_task_id = %task.id,
+ "Created red team task for contract"
+ );
+
+ // Find a daemon to run the red team task
+ for entry in state.daemon_connections.iter() {
+ let daemon = entry.value();
+ if daemon.owner_id == owner_id {
+ // Update task with daemon assignment
+ let update_req = UpdateTaskRequest {
+ status: Some("starting".to_string()),
+ daemon_id: Some(daemon.id),
+ version: Some(task.version),
+ ..Default::default()
+ };
+
+ match repository::update_task_for_owner(pool, task.id, owner_id, update_req).await {
+ Ok(Some(updated_task)) => {
+ // Send spawn command to daemon
+ let cmd = DaemonCommand::SpawnTask {
+ task_id: updated_task.id,
+ task_name: updated_task.name.clone(),
+ plan: updated_task.plan.clone(),
+ repo_url: None, // Red team doesn't need a repo
+ base_branch: None,
+ target_branch: None,
+ parent_task_id: None,
+ depth: 0,
+ is_orchestrator: false,
+ target_repo_path: None,
+ completion_action: None,
+ continue_from_task_id: None,
+ copy_files: None,
+ contract_id: Some(contract_id),
+ is_supervisor: false,
+ autonomous_loop: false,
+ resume_session: false,
+ conversation_history: None,
+ patch_data: None,
+ patch_base_sha: None,
+ local_only: true, // Red team is always local-only
+ };
+
+ if let Err(e) = state.send_daemon_command(daemon.id, cmd).await {
+ tracing::warn!(
+ error = %e,
+ daemon_id = %daemon.id,
+ red_team_task_id = %task.id,
+ "Failed to send red team spawn command"
+ );
+ // Rollback
+ let rollback_req = UpdateTaskRequest {
+ status: Some("pending".to_string()),
+ clear_daemon_id: true,
+ ..Default::default()
+ };
+ let _ = repository::update_task_for_owner(pool, task.id, owner_id, rollback_req).await;
+ } else {
+ tracing::info!(
+ red_team_task_id = %task.id,
+ daemon_id = %daemon.id,
+ "Red team task spawn command sent"
+ );
+ return Ok(updated_task);
+ }
+ }
+ Ok(None) => {
+ tracing::warn!(red_team_task_id = %task.id, "Red team task not found when updating daemon_id");
+ }
+ Err(e) => {
+ tracing::error!(red_team_task_id = %task.id, error = %e, "Failed to update red team task with daemon_id");
+ }
+ }
+ break;
+ }
+ }
+
+ // Return the task even if we couldn't start it on a daemon
+ // It will remain pending and can be started later
+ Ok(task)
+}
diff --git a/makima/src/server/handlers/mod.rs b/makima/src/server/handlers/mod.rs
index 609b63b..b496922 100644
--- a/makima/src/server/handlers/mod.rs
+++ b/makima/src/server/handlers/mod.rs
@@ -13,6 +13,7 @@ pub mod mesh;
pub mod mesh_chat;
pub mod mesh_daemon;
pub mod mesh_merge;
+pub mod mesh_red_team;
pub mod mesh_supervisor;
pub mod mesh_ws;
pub mod repository_history;
diff --git a/makima/src/server/handlers/transcript_analysis.rs b/makima/src/server/handlers/transcript_analysis.rs
index 8eb50c7..3c283da 100644
--- a/makima/src/server/handlers/transcript_analysis.rs
+++ b/makima/src/server/handlers/transcript_analysis.rs
@@ -279,6 +279,8 @@ pub async fn create_contract_from_analysis(
autonomous_loop: None,
phase_guard: None,
local_only: None,
+ red_team_enabled: None,
+ red_team_prompt: None,
};
let contract = match repository::create_contract_for_owner(pool, auth.owner_id, contract_req).await {
@@ -358,6 +360,7 @@ pub async fn create_contract_from_analysis(
continue_from_task_id: None,
copy_files: None,
is_supervisor: false,
+ is_red_team: false,
checkpoint_sha: None,
priority: match item.priority.as_deref() {
Some("high") => 10,
@@ -531,6 +534,7 @@ pub async fn update_contract_from_analysis(
continue_from_task_id: None,
copy_files: None,
is_supervisor: false,
+ is_red_team: false,
checkpoint_sha: None,
priority: 0,
merge_mode: None,