meerkat-core 0.4.6

Core agent logic for Meerkat (no I/O deps)
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
//! Agent runner interface.

use crate::budget::Budget;
use crate::error::AgentError;
use crate::event::{AgentEvent, ScopedAgentEvent, StreamScopeFrame};
use crate::hooks::{HookDecision, HookInvocation, HookPatch, HookPoint};
use crate::ops::{
    ForkBranch, ForkBudgetPolicy, OperationId, OperationResult, SpawnSpec, ToolAccessPolicy,
};
use crate::retry::RetryPolicy;
use crate::service::TurnToolOverlay;
use crate::session::Session;
use crate::state::LoopState;
#[cfg(target_arch = "wasm32")]
use crate::tokio;
use crate::tool_scope::{
    EXTERNAL_TOOL_FILTER_METADATA_KEY, ToolFilter, ToolScopeRevision, ToolScopeStageError,
};
use crate::types::{Message, RunResult};
use async_trait::async_trait;
use std::collections::HashSet;
use std::sync::Arc;
use tokio::sync::mpsc;

use super::{
    Agent, AgentBuilder, AgentLlmClient, AgentSessionStore, AgentToolDispatcher,
    FilteredToolDispatcher,
};

fn spawn_scoped_forwarder(
    mut child_event_rx: mpsc::Receiver<AgentEvent>,
    scoped_tx: mpsc::Sender<ScopedAgentEvent>,
    parent_scope_path: Arc<Vec<StreamScopeFrame>>,
    child_scope_frame: StreamScopeFrame,
) -> tokio::task::JoinHandle<()> {
    let base_scope_path = if parent_scope_path.is_empty() {
        vec![
            StreamScopeFrame::Primary {
                session_id: "unknown".to_string(),
            },
            child_scope_frame,
        ]
    } else {
        let mut path = (*parent_scope_path).clone();
        path.push(child_scope_frame);
        path
    };

    tokio::spawn(async move {
        while let Some(event) = child_event_rx.recv().await {
            let scoped = ScopedAgentEvent::new(base_scope_path.clone(), event);
            if scoped_tx.send(scoped).await.is_err() {
                break;
            }
        }
    })
}

/// Minimal runner interface for an Agent.
#[cfg_attr(target_arch = "wasm32", async_trait(?Send))]
#[cfg_attr(not(target_arch = "wasm32"), async_trait)]
pub trait AgentRunner: Send {
    async fn run(&mut self, prompt: String) -> Result<RunResult, AgentError>;

    async fn run_with_events(
        &mut self,
        prompt: String,
        tx: mpsc::Sender<AgentEvent>,
    ) -> Result<RunResult, AgentError>;
}

impl<C, T, S> Agent<C, T, S>
where
    C: AgentLlmClient + ?Sized,
    T: AgentToolDispatcher + ?Sized,
    S: AgentSessionStore + ?Sized,
{
    /// Stage an external tool visibility filter update for subsequent turns.
    pub fn stage_external_tool_filter(
        &mut self,
        filter: ToolFilter,
    ) -> Result<ToolScopeRevision, ToolScopeStageError> {
        let handle = self.tool_scope.handle();
        let revision = handle.stage_external_filter(filter.clone())?;
        let _ = handle.staged_revision();
        if let Ok(value) = serde_json::to_value(filter) {
            self.session
                .set_metadata(EXTERNAL_TOOL_FILTER_METADATA_KEY, value);
        }
        Ok(revision)
    }

    /// Set or clear a per-turn flow tool overlay.
    pub fn set_flow_tool_overlay(
        &mut self,
        overlay: Option<TurnToolOverlay>,
    ) -> Result<(), ToolScopeStageError> {
        let handle = self.tool_scope.handle();
        if let Some(overlay) = overlay {
            let allow = overlay
                .allowed_tools
                .map(|tools| tools.into_iter().collect::<HashSet<_>>());
            let deny = overlay
                .blocked_tools
                .unwrap_or_default()
                .into_iter()
                .collect::<HashSet<_>>();
            handle.set_turn_overlay(allow, deny)?;
        } else {
            handle.clear_turn_overlay();
        }
        Ok(())
    }

    #[cfg(test)]
    pub(crate) fn inject_tool_scope_boundary_failure_once_for_test(&self) {
        self.tool_scope.inject_boundary_failure_once_for_test();
    }
}

impl<C, T, S> Agent<C, T, S>
where
    C: AgentLlmClient + ?Sized + 'static,
    T: AgentToolDispatcher + ?Sized + 'static,
    S: AgentSessionStore + ?Sized + 'static,
{
    /// Create a new agent builder
    pub fn builder() -> AgentBuilder {
        AgentBuilder::new()
    }

    /// Get the current session
    pub fn session(&self) -> &Session {
        &self.session
    }

    /// Get mutable access to the session (for setting metadata)
    pub fn session_mut(&mut self) -> &mut Session {
        &mut self.session
    }

    /// Get the current budget
    pub fn budget(&self) -> &Budget {
        &self.budget
    }

    /// Get the current state
    pub fn state(&self) -> &LoopState {
        &self.state
    }

    /// Get the retry policy
    pub fn retry_policy(&self) -> &RetryPolicy {
        &self.retry_policy
    }

    /// Get the current nesting depth
    pub fn depth(&self) -> u32 {
        self.depth
    }

    /// Get the event tap for interaction-scoped streaming.
    pub fn event_tap(&self) -> &crate::event_tap::EventTap {
        &self.event_tap
    }

    /// Get shared runtime system-context control state.
    pub fn system_context_state(
        &self,
    ) -> Arc<std::sync::Mutex<crate::session::SessionSystemContextState>> {
        Arc::clone(&self.system_context_state)
    }

    /// Clone the current session with the latest shared system-context state merged into metadata.
    pub fn session_with_system_context_state(&self) -> Session {
        let mut session = self.session.clone();
        let state = match self.system_context_state.lock() {
            Ok(guard) => guard.clone(),
            Err(poisoned) => {
                tracing::warn!("system-context state lock poisoned while cloning session");
                poisoned.into_inner().clone()
            }
        };
        if let Err(err) = session.set_system_context_state(state) {
            tracing::warn!(error = %err, "failed to serialize system-context state into session");
        }
        session
    }

    /// Synchronize the shared system-context state into the in-memory session metadata.
    pub(crate) fn sync_system_context_state_to_session(&mut self) {
        let state = match self.system_context_state.lock() {
            Ok(guard) => guard.clone(),
            Err(poisoned) => {
                tracing::warn!("system-context state lock poisoned while syncing session");
                poisoned.into_inner().clone()
            }
        };
        if let Err(err) = self.session.set_system_context_state(state) {
            tracing::warn!(error = %err, "failed to serialize system-context state into session");
        }
    }

    /// Apply all pending system-context appends at the current LLM boundary.
    pub(crate) fn apply_pending_system_context_boundary(&mut self) -> usize {
        let pending = {
            let mut state = match self.system_context_state.lock() {
                Ok(guard) => guard,
                Err(poisoned) => {
                    tracing::warn!("system-context state lock poisoned while applying boundary");
                    poisoned.into_inner()
                }
            };
            if state.pending.is_empty() {
                return 0;
            }
            let pending = state.pending.clone();
            state.mark_pending_applied();
            pending
        };

        self.session.append_system_context_blocks(&pending);
        self.sync_system_context_state_to_session();
        pending.len()
    }

    /// Persist the current session through the configured checkpointer after syncing control state.
    pub(crate) async fn checkpoint_current_session(&mut self) {
        self.sync_system_context_state_to_session();
        if let Some(ref cp) = self.checkpointer {
            cp.checkpoint(&self.session).await;
        }
    }

    /// Spawn a new sub-agent with minimal context
    ///
    /// The sub-agent runs independently with its own budget and tool access.
    /// Results are collected at turn boundaries.
    pub async fn spawn(&self, spec: SpawnSpec) -> Result<OperationId, AgentError> {
        // Check depth limit
        if self.depth + 1 > self.sub_agent_manager.limits.max_depth {
            return Err(AgentError::DepthLimitExceeded {
                depth: self.depth + 1,
                max: self.sub_agent_manager.limits.max_depth,
            });
        }

        // Check if we can spawn more sub-agents
        if !self.sub_agent_manager.can_spawn().await {
            return Err(AgentError::SubAgentLimitExceeded {
                limit: self.sub_agent_manager.limits.max_concurrent_agents,
            });
        }

        // Validate tool access policy
        let all_tools = self.tools.tools();
        let allowed_tools = self
            .sub_agent_manager
            .apply_tool_access_policy(&all_tools, &spec.tool_access);

        if let ToolAccessPolicy::AllowList(ref names) = spec.tool_access {
            for name in names {
                if !all_tools.iter().any(|t| &t.name == name) {
                    return Err(AgentError::InvalidToolAccess { tool: name.clone() });
                }
            }
        }

        // Apply context strategy to get messages for sub-agent
        let messages = self
            .sub_agent_manager
            .apply_context_strategy(&self.session, &spec.context);

        // Create sub-agent session with context
        let mut sub_session = Session::new();
        for msg in messages {
            sub_session.push(msg);
        }
        if let Some(sys_prompt) = &spec.system_prompt {
            sub_session.set_system_prompt(sys_prompt.clone());
        }

        // Generate operation ID
        let op_id = OperationId::new();

        // Register the sub-agent
        self.sub_agent_manager
            .register(op_id.clone(), "spawn".to_string())
            .await?;

        // Clone components for the spawned task
        let client = self.client.clone();
        let store = self.store.clone();
        let prompt = spec.prompt.clone();
        let budget = spec.budget.clone();
        let sub_agent_manager = self.sub_agent_manager.clone();
        let op_id_clone = op_id.clone();
        let depth = self.depth + 1;
        let model = self.config.model.clone();
        let max_tokens = self.config.max_tokens_per_turn;
        let parent_scoped_event_tx = self.default_scoped_event_tx.clone();
        let parent_scope_path = self.default_scope_path.clone();

        // Create filtered tools based on policy
        let allowed_tool_names: Vec<String> =
            allowed_tools.iter().map(|t| t.name.clone()).collect();
        let filtered_tools = Arc::new(FilteredToolDispatcher::new(
            self.tools.clone(),
            allowed_tool_names,
        ));

        // Spawn the sub-agent in a background task
        tokio::spawn(async move {
            let start = crate::time_compat::Instant::now();

            // Build sub-agent with filtered tools
            let mut sub_agent = AgentBuilder::new()
                .model(&model)
                .max_tokens_per_turn(max_tokens)
                .budget(budget)
                .resume_session(sub_session)
                .build(client, filtered_tools, store)
                .await;

            let (result, forwarder_task) = if let Some(scoped_tx) = parent_scoped_event_tx {
                let (child_event_tx, child_event_rx) = mpsc::channel::<AgentEvent>(64);
                let child_scope_frame = StreamScopeFrame::SubAgent {
                    agent_id: op_id_clone.to_string(),
                    tool_call_id: None,
                    label: Some("spawn".to_string()),
                };
                let forwarder = spawn_scoped_forwarder(
                    child_event_rx,
                    scoped_tx,
                    Arc::new(parent_scope_path.clone()),
                    child_scope_frame,
                );
                (
                    sub_agent.run_with_events(prompt, child_event_tx).await,
                    Some(forwarder),
                )
            } else {
                (sub_agent.run(prompt).await, None)
            };
            if let Some(forwarder) = forwarder_task {
                let _ = forwarder.await;
            }

            // Report completion
            match result {
                Ok(run_result) => {
                    sub_agent_manager
                        .complete(
                            &op_id_clone,
                            OperationResult {
                                id: op_id_clone.clone(),
                                content: run_result.text,
                                is_error: false,
                                duration_ms: start.elapsed().as_millis() as u64,
                                tokens_used: run_result.usage.total_tokens(),
                            },
                        )
                        .await;
                }
                Err(e) => {
                    sub_agent_manager.fail(&op_id_clone, e.to_string()).await;
                }
            }
        });

        tracing::info!(
            "Spawned sub-agent {} at depth {} with {} tools",
            op_id,
            depth,
            allowed_tools.len()
        );

        Ok(op_id)
    }

    /// Fork the current conversation into parallel branches
    ///
    /// Each branch gets a copy of the full conversation history and runs independently.
    pub async fn fork(
        &self,
        branches: Vec<ForkBranch>,
        budget_policy: ForkBudgetPolicy,
    ) -> Result<Vec<OperationId>, AgentError> {
        // Check depth limit
        if self.depth + 1 > self.sub_agent_manager.limits.max_depth {
            return Err(AgentError::DepthLimitExceeded {
                depth: self.depth + 1,
                max: self.sub_agent_manager.limits.max_depth,
            });
        }

        // Check if we can spawn enough sub-agents
        let running = self.sub_agent_manager.running_ids().await.len();
        if running + branches.len() > self.sub_agent_manager.limits.max_concurrent_agents {
            return Err(AgentError::SubAgentLimitExceeded {
                limit: self.sub_agent_manager.limits.max_concurrent_agents,
            });
        }

        // Allocate budget for each branch
        let remaining_tokens = self.budget.remaining();
        let budgets = self.sub_agent_manager.allocate_fork_budget(
            remaining_tokens,
            branches.len(),
            &budget_policy,
        );

        let mut op_ids = Vec::with_capacity(branches.len());

        for (i, branch) in branches.into_iter().enumerate() {
            let op_id = OperationId::new();

            // Validate tool access if specified
            if let Some(ToolAccessPolicy::AllowList(names)) = &branch.tool_access {
                let all_tools = self.tools.tools();
                for name in names {
                    if !all_tools.iter().any(|t| &t.name == name) {
                        return Err(AgentError::InvalidToolAccess { tool: name.clone() });
                    }
                }
            }

            // Register the branch as a sub-agent
            self.sub_agent_manager
                .register(op_id.clone(), branch.name.clone())
                .await?;

            // Apply tool access policy for this branch
            let all_tools = self.tools.tools();
            let allowed_tools = match &branch.tool_access {
                Some(policy) => self
                    .sub_agent_manager
                    .apply_tool_access_policy(&all_tools, policy),
                None => all_tools.to_vec(), // Inherit all tools
            };
            let allowed_tool_names: Vec<String> =
                allowed_tools.iter().map(|t| t.name.clone()).collect();
            let filtered_tools = Arc::new(FilteredToolDispatcher::new(
                self.tools.clone(),
                allowed_tool_names,
            ));

            // Clone components for the spawned task
            let client = self.client.clone();
            let store = self.store.clone();
            let prompt = branch.prompt.clone();
            let budget = budgets[i].clone();
            let sub_agent_manager = self.sub_agent_manager.clone();
            let op_id_clone = op_id.clone();
            let model = self.config.model.clone();
            let max_tokens = self.config.max_tokens_per_turn;
            let branch_name = branch.name.clone();
            let parent_scoped_event_tx = self.default_scoped_event_tx.clone();
            let parent_scope_path = self.default_scope_path.clone();

            // Create session with full history (fork uses FullHistory context)
            let mut fork_session = Session::new();
            for msg in self.session.messages() {
                fork_session.push(msg.clone());
            }

            // Spawn the branch in a background task
            tokio::spawn(async move {
                let start = crate::time_compat::Instant::now();

                // Build sub-agent for this branch with filtered tools
                let mut sub_agent = AgentBuilder::new()
                    .model(&model)
                    .max_tokens_per_turn(max_tokens)
                    .budget(budget)
                    .resume_session(fork_session)
                    .build(client, filtered_tools, store)
                    .await;

                let (result, forwarder_task) = if let Some(scoped_tx) = parent_scoped_event_tx {
                    let (child_event_tx, child_event_rx) = mpsc::channel::<AgentEvent>(64);
                    let child_scope_frame = StreamScopeFrame::SubAgent {
                        agent_id: op_id_clone.to_string(),
                        tool_call_id: None,
                        label: Some(branch_name.clone()),
                    };
                    let forwarder = spawn_scoped_forwarder(
                        child_event_rx,
                        scoped_tx,
                        Arc::new(parent_scope_path.clone()),
                        child_scope_frame,
                    );
                    (
                        sub_agent.run_with_events(prompt, child_event_tx).await,
                        Some(forwarder),
                    )
                } else {
                    (sub_agent.run(prompt).await, None)
                };
                if let Some(forwarder) = forwarder_task {
                    let _ = forwarder.await;
                }

                // Report completion
                match result {
                    Ok(run_result) => {
                        sub_agent_manager
                            .complete(
                                &op_id_clone,
                                OperationResult {
                                    id: op_id_clone.clone(),
                                    content: format!("[{}] {}", branch_name, run_result.text),
                                    is_error: false,
                                    duration_ms: start.elapsed().as_millis() as u64,
                                    tokens_used: run_result.usage.total_tokens(),
                                },
                            )
                            .await;
                    }
                    Err(e) => {
                        sub_agent_manager.fail(&op_id_clone, e.to_string()).await;
                    }
                }
            });

            tracing::info!(
                "Forked branch '{}' as {} at depth {}",
                branch.name,
                op_id,
                self.depth + 1
            );

            op_ids.push(op_id);
        }

        Ok(op_ids)
    }

    /// Cancel a running sub-agent
    pub async fn cancel_sub_agent(&self, op_id: &OperationId) {
        self.sub_agent_manager.cancel(op_id).await;
    }

    /// Collect completed sub-agent results (called at turn boundaries)
    pub async fn collect_sub_agent_results(&self) -> Vec<OperationResult> {
        self.sub_agent_manager.collect_completed().await
    }

    /// Check if there are running sub-agents
    pub async fn has_running_sub_agents(&self) -> bool {
        self.sub_agent_manager.has_running().await
    }

    async fn run_started_hooks(
        &self,
        prompt: &str,
        event_tx: Option<&mpsc::Sender<AgentEvent>>,
    ) -> Result<(), AgentError> {
        let report = self
            .execute_hooks(
                HookInvocation {
                    point: HookPoint::RunStarted,
                    session_id: self.session.id().clone(),
                    turn_number: None,
                    prompt: Some(prompt.to_string()),
                    error: None,
                    llm_request: None,
                    llm_response: None,
                    tool_call: None,
                    tool_result: None,
                },
                event_tx,
            )
            .await?;

        if let Some(HookDecision::Deny {
            reason_code,
            message,
            payload,
            ..
        }) = report.decision
        {
            return Err(AgentError::HookDenied {
                point: HookPoint::RunStarted,
                reason_code,
                message,
                payload,
            });
        }
        Ok(())
    }

    async fn run_completed_hooks(
        &mut self,
        result: &mut RunResult,
        event_tx: Option<&mpsc::Sender<AgentEvent>>,
    ) -> Result<(), AgentError> {
        let report = self
            .execute_hooks(
                HookInvocation {
                    point: HookPoint::RunCompleted,
                    session_id: self.session.id().clone(),
                    turn_number: Some(result.turns),
                    prompt: None,
                    error: None,
                    llm_request: None,
                    llm_response: None,
                    tool_call: None,
                    tool_result: None,
                },
                event_tx,
            )
            .await?;

        if let Some(HookDecision::Deny {
            reason_code,
            message,
            payload,
            ..
        }) = report.decision
        {
            return Err(AgentError::HookDenied {
                point: HookPoint::RunCompleted,
                reason_code,
                message,
                payload,
            });
        }

        for outcome in &report.outcomes {
            for patch in &outcome.patches {
                if let HookPatch::RunResult { text } = patch {
                    crate::event_tap::tap_emit(
                        &self.event_tap,
                        event_tx,
                        AgentEvent::HookRewriteApplied {
                            hook_id: outcome.hook_id.to_string(),
                            point: HookPoint::RunCompleted,
                            patch: HookPatch::RunResult { text: text.clone() },
                        },
                    )
                    .await;
                    result.text.clone_from(text);
                    self.apply_run_result_text_patch(text);
                }
            }
        }
        if let Err(err) = self.store.save(&self.session).await {
            tracing::warn!("Failed to save session after run_completed hooks: {}", err);
        }
        Ok(())
    }

    fn apply_run_result_text_patch(&mut self, text: &str) {
        use super::state::rewrite_assistant_text;
        let messages = self.session.messages_mut();
        if let Some(last_assistant) = messages
            .iter_mut()
            .rev()
            .find(|message| matches!(message, Message::BlockAssistant(_) | Message::Assistant(_)))
        {
            match last_assistant {
                Message::BlockAssistant(block_assistant) => {
                    rewrite_assistant_text(&mut block_assistant.blocks, text.to_string());
                }
                Message::Assistant(assistant) => {
                    assistant.content = text.to_string();
                }
                _ => {}
            }
            self.session.touch();
        }
    }

    async fn run_failed_hooks(
        &self,
        error: &AgentError,
        event_tx: Option<&mpsc::Sender<AgentEvent>>,
    ) -> Result<(), AgentError> {
        let report = self
            .execute_hooks(
                HookInvocation {
                    point: HookPoint::RunFailed,
                    session_id: self.session.id().clone(),
                    turn_number: None,
                    prompt: None,
                    error: Some(error.to_string()),
                    llm_request: None,
                    llm_response: None,
                    tool_call: None,
                    tool_result: None,
                },
                event_tx,
            )
            .await?;

        if let Some(HookDecision::Deny {
            reason_code,
            message,
            payload,
            ..
        }) = report.decision
        {
            return Err(AgentError::HookDenied {
                point: HookPoint::RunFailed,
                reason_code,
                message,
                payload,
            });
        }
        Ok(())
    }

    /// Run the agent with a user message.
    pub async fn run(&mut self, user_input: String) -> Result<RunResult, AgentError> {
        self.run_inner(user_input, None).await
    }

    /// Run the agent with events streamed to the provided channel.
    pub async fn run_with_events(
        &mut self,
        user_input: String,
        event_tx: mpsc::Sender<AgentEvent>,
    ) -> Result<RunResult, AgentError> {
        self.run_inner(user_input, Some(event_tx)).await
    }

    /// Run the agent using the pending user message already in the session.
    ///
    /// This is useful when the session has been pre-populated with a user message
    /// (e.g., via `create_spawn_session` or `create_fork_session`). Unlike `run()`,
    /// this method does NOT add a new user message — it runs directly from the
    /// session's current state.
    ///
    /// Returns an error if the session doesn't have a pending user message.
    pub async fn run_pending(&mut self) -> Result<RunResult, AgentError> {
        self.run_pending_inner(None).await
    }

    /// Run the agent using the pending user message, with event streaming.
    ///
    /// Like `run_pending()`, but emits events to the provided channel.
    pub async fn run_pending_with_events(
        &mut self,
        event_tx: mpsc::Sender<AgentEvent>,
    ) -> Result<RunResult, AgentError> {
        self.run_pending_inner(Some(event_tx)).await
    }

    /// Core run implementation shared by `run()` and `run_with_events()`.
    ///
    /// Adds user_input as a user message, emits lifecycle events when `event_tx`
    /// is provided, and delegates to `run_loop`.
    async fn run_inner(
        &mut self,
        user_input: String,
        event_tx: Option<mpsc::Sender<AgentEvent>>,
    ) -> Result<RunResult, AgentError> {
        let event_tx = event_tx.or_else(|| self.default_event_tx.clone());

        // Reset state for new run (allows multi-turn on same agent)
        self.state = LoopState::CallingLlm;

        // Apply canonical per-turn skill references staged by the surface.
        let user_input = self.apply_skill_ref(user_input).await;

        let run_prompt = user_input.clone();

        // Add user message
        self.session.push(Message::User(crate::types::UserMessage {
            content: user_input,
        }));

        if let Some(ref tx) = event_tx {
            let _ = crate::event_tap::tap_emit(
                &self.event_tap,
                Some(tx),
                AgentEvent::RunStarted {
                    session_id: self.session.id().clone(),
                    prompt: run_prompt.clone(),
                },
            )
            .await;
        }

        self.run_started_hooks(&run_prompt, event_tx.as_ref())
            .await?;

        match self.run_loop(event_tx.clone()).await {
            Ok(mut result) => {
                self.run_completed_hooks(&mut result, event_tx.as_ref())
                    .await?;
                Ok(result)
            }
            Err(err) => {
                if let Err(hook_err) = self.run_failed_hooks(&err, event_tx.as_ref()).await {
                    tracing::warn!(?hook_err, "run_failed hook execution failed");
                }
                if let Some(ref tx) = event_tx {
                    let _ = crate::event_tap::tap_emit(
                        &self.event_tap,
                        Some(tx),
                        AgentEvent::RunFailed {
                            session_id: self.session.id().clone(),
                            error: err.to_string(),
                        },
                    )
                    .await;
                }
                Err(err)
            }
        }
    }

    /// Core run-pending implementation shared by `run_pending()` and
    /// `run_pending_with_events()`.
    ///
    /// Uses the existing pending user message in the session (does NOT push a new one).
    /// Emits lifecycle events when `event_tx` is provided.
    async fn run_pending_inner(
        &mut self,
        event_tx: Option<mpsc::Sender<AgentEvent>>,
    ) -> Result<RunResult, AgentError> {
        let event_tx = event_tx.or_else(|| self.default_event_tx.clone());

        let pending_prompt = self.session.messages().last().and_then(|m| match m {
            Message::User(u) => Some(u.content.clone()),
            _ => None,
        });

        let Some(prompt) = pending_prompt else {
            return Err(AgentError::ConfigError(
                "run_pending requires a pending user message in the session".to_string(),
            ));
        };

        // Reset state for new run (allows multi-turn on same agent)
        self.state = LoopState::CallingLlm;

        if let Some(ref tx) = event_tx {
            let _ = crate::event_tap::tap_emit(
                &self.event_tap,
                Some(tx),
                AgentEvent::RunStarted {
                    session_id: self.session.id().clone(),
                    prompt: prompt.clone(),
                },
            )
            .await;
        }

        self.run_started_hooks(&prompt, event_tx.as_ref()).await?;

        match self.run_loop(event_tx.clone()).await {
            Ok(mut result) => {
                self.run_completed_hooks(&mut result, event_tx.as_ref())
                    .await?;
                Ok(result)
            }
            Err(err) => {
                if let Err(hook_err) = self.run_failed_hooks(&err, event_tx.as_ref()).await {
                    tracing::warn!(?hook_err, "run_failed hook execution failed");
                }
                if let Some(ref tx) = event_tx {
                    let _ = crate::event_tap::tap_emit(
                        &self.event_tap,
                        Some(tx),
                        AgentEvent::RunFailed {
                            session_id: self.session.id().clone(),
                            error: err.to_string(),
                        },
                    )
                    .await;
                }
                Err(err)
            }
        }
    }

    /// Cancel the current run
    pub fn cancel(&mut self) {
        if !self.state.is_terminal() {
            let _ = self.state.transition(LoopState::Cancelling);
        }
    }

    /// Consume canonical pending `skill_references` staged by the surface and
    /// prepend resolved skill bodies to the next user input.
    ///
    /// Compatibility slash refs are handled at transport/resolver boundaries;
    /// core runtime no longer parses slash refs directly.
    async fn apply_skill_ref(&mut self, user_input: String) -> String {
        let engine = match &self.skill_engine {
            Some(e) => e.clone(),
            None => return user_input,
        };

        let mut prefix_parts: Vec<String> = Vec::new();

        // 1. Consume pending_skill_references (from wire format / API)
        if let Some(refs) = self.pending_skill_references.take()
            && !refs.is_empty()
        {
            let canonical_ids: Vec<crate::skills::SkillId> = refs
                .into_iter()
                .map(|key| {
                    crate::skills::SkillId(format!("{}/{}", key.source_uuid, key.skill_name))
                })
                .collect();
            match engine.resolve_and_render(&canonical_ids).await {
                Ok(resolved) => {
                    for skill in &resolved {
                        tracing::info!(
                            skill_id = %skill.id.0,
                            "Per-turn skill activation via skill_references"
                        );
                        prefix_parts.push(skill.rendered_body.clone());
                    }
                }
                Err(e) => {
                    tracing::warn!(
                        error = %e,
                        "Failed to resolve source-pinned skill_references"
                    );
                }
            }
        }

        if prefix_parts.is_empty() {
            return user_input;
        }

        if user_input.is_empty() {
            prefix_parts.join("\n\n")
        } else {
            format!("{}\n\n{user_input}", prefix_parts.join("\n\n"))
        }
    }
}