kumiho-construct 2026.5.11

Construct — memory-native AI agent runtime powered by Kumiho
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
//! Interactive approval workflow for supervised mode.
//!
//! Provides a pre-execution hook that prompts the user before tool calls,
//! with session-scoped "Always" allowlists and audit logging.

use crate::config::AutonomyConfig;
use crate::security::AutonomyLevel;
use crate::trust::{CorrectionType, TrustTracker};
use chrono::Utc;
use parking_lot::Mutex;
use serde::{Deserialize, Serialize};
use std::collections::HashSet;
use std::io::{self, BufRead, Write};
use std::sync::Arc;

// ── Types ────────────────────────────────────────────────────────

/// A request to approve a tool call before execution.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ApprovalRequest {
    pub tool_name: String,
    pub arguments: serde_json::Value,
}

/// The user's response to an approval request.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize)]
#[serde(rename_all = "lowercase")]
pub enum ApprovalResponse {
    /// Execute this one call.
    Yes,
    /// Deny this call.
    No,
    /// Execute and add tool to session-scoped allowlist.
    Always,
}

/// A single audit log entry for an approval decision.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ApprovalLogEntry {
    pub timestamp: String,
    pub tool_name: String,
    pub arguments_summary: String,
    pub decision: ApprovalResponse,
    pub channel: String,
}

// ── ApprovalManager ──────────────────────────────────────────────

/// Manages the approval workflow for tool calls.
///
/// - Checks config-level `auto_approve` / `always_ask` lists
/// - Maintains a session-scoped "always" allowlist
/// - Records an audit trail of all decisions
///
/// Two modes:
/// - **Interactive** (CLI): tools needing approval trigger a stdin prompt.
/// - **Non-interactive** (channels): tools needing approval are auto-denied
///   because there is no interactive operator to approve them. `auto_approve`
///   policy is still enforced, and `always_ask` / supervised-default tools are
///   denied rather than silently allowed.
pub struct ApprovalManager {
    /// Tools that never need approval (from config).
    auto_approve: HashSet<String>,
    /// Tools that always need approval, ignoring session allowlist.
    always_ask: HashSet<String>,
    /// Autonomy level from config.
    autonomy_level: AutonomyLevel,
    /// When `true`, tools that would require interactive approval are
    /// auto-denied instead. Used for channel-driven (non-CLI) runs.
    non_interactive: bool,
    /// Session-scoped allowlist built from "Always" responses.
    session_allowlist: Mutex<HashSet<String>>,
    /// Audit trail of approval decisions.
    audit_log: Mutex<Vec<ApprovalLogEntry>>,
    /// Optional per-domain trust tracker. When present, a domain in regression
    /// downgrades the effective autonomy level by one tier in `needs_approval`.
    /// Denied approvals are recorded as `UserOverride` corrections; callers
    /// can also report successes and other corrections via the public methods.
    trust: Option<Arc<Mutex<TrustTracker>>>,
}

impl ApprovalManager {
    /// Create an interactive (CLI) approval manager from autonomy config.
    pub fn from_config(config: &AutonomyConfig) -> Self {
        Self {
            auto_approve: config.auto_approve.iter().cloned().collect(),
            always_ask: config.always_ask.iter().cloned().collect(),
            autonomy_level: config.level,
            non_interactive: false,
            session_allowlist: Mutex::new(HashSet::new()),
            audit_log: Mutex::new(Vec::new()),
            trust: None,
        }
    }

    /// Create a non-interactive approval manager for channel-driven runs.
    ///
    /// Enforces the same `auto_approve` / `always_ask` / supervised policies
    /// as the CLI manager, but tools that would require interactive approval
    /// are auto-denied instead of prompting (since there is no operator).
    pub fn for_non_interactive(config: &AutonomyConfig) -> Self {
        Self {
            auto_approve: config.auto_approve.iter().cloned().collect(),
            always_ask: config.always_ask.iter().cloned().collect(),
            autonomy_level: config.level,
            non_interactive: true,
            session_allowlist: Mutex::new(HashSet::new()),
            audit_log: Mutex::new(Vec::new()),
            trust: None,
        }
    }

    /// Attach a trust tracker. Denials recorded as `UserOverride` corrections;
    /// regression on a tool's domain downgrades effective autonomy one tier.
    pub fn with_trust_tracker(mut self, tracker: Arc<Mutex<TrustTracker>>) -> Self {
        self.trust = Some(tracker);
        self
    }

    /// Current effective autonomy level for `domain`, applying trust regression
    /// downgrade when a tracker is attached. `domain` is typically the tool name.
    fn effective_autonomy(&self, domain: &str) -> AutonomyLevel {
        match &self.trust {
            Some(trust) => trust
                .lock()
                .effective_autonomy_level(domain, self.autonomy_level),
            None => self.autonomy_level,
        }
    }

    /// Record a successful tool execution against the trust tracker, if any.
    /// Safe no-op when no tracker is attached.
    pub fn record_tool_success(&self, tool_name: &str) {
        if let Some(trust) = &self.trust {
            trust.lock().record_success(tool_name);
        }
    }

    /// Record a correction (quality failure, SOP deviation, etc.) against the
    /// trust tracker, if any. Safe no-op when no tracker is attached.
    pub fn record_tool_correction(
        &self,
        tool_name: &str,
        correction_type: CorrectionType,
        description: &str,
    ) {
        if let Some(trust) = &self.trust {
            trust
                .lock()
                .record_correction(tool_name, correction_type, description);
        }
    }

    /// Returns `true` when this manager operates in non-interactive mode
    /// (i.e. for channel-driven runs where no operator can approve).
    pub fn is_non_interactive(&self) -> bool {
        self.non_interactive
    }

    /// Check whether a tool call requires interactive approval.
    ///
    /// Returns `true` if the call needs a prompt, `false` if it can proceed.
    pub fn needs_approval(&self, tool_name: &str) -> bool {
        let level = self.effective_autonomy(tool_name);

        // Full autonomy never prompts.
        if level == AutonomyLevel::Full {
            return false;
        }

        if level == AutonomyLevel::ReadOnly {
            // GLOBAL ReadOnly: the call is blocked downstream by
            // SecurityPolicy::can_act — no prompt needed.
            if self.autonomy_level == AutonomyLevel::ReadOnly {
                return false;
            }
            // Per-tool regression dropped us here while global autonomy is
            // Supervised or Full. SecurityPolicy::can_act only inspects the
            // global level and will let the call proceed, so without a
            // prompt the denied tool would execute silently. Force a prompt
            // (or auto-deny in non-interactive mode via the caller).
            return true;
        }

        // always_ask overrides everything.
        if self.always_ask.contains("*") || self.always_ask.contains(tool_name) {
            return true;
        }

        // Channel-driven shell execution is still guarded by the shell tool's
        // own command allowlist and risk policy. Skipping the outer approval
        // gate here lets low-risk allowlisted commands (e.g. `ls`) work in
        // non-interactive channels without silently allowing medium/high-risk
        // commands.
        if self.non_interactive && tool_name == "shell" {
            return false;
        }

        // MCP-namespaced tools (e.g. "kumiho-memory__kumiho_memory_engage")
        // are injected by admin-configured MCP servers. In non-interactive
        // mode, auto-approve them — they were explicitly provisioned and
        // there is no operator to prompt.
        if self.non_interactive && tool_name.contains("__") {
            return false;
        }

        // auto_approve skips the prompt.
        if self.auto_approve.contains("*") || self.auto_approve.contains(tool_name) {
            return false;
        }

        // Session allowlist (from prior "Always" responses).
        let allowlist = self.session_allowlist.lock();
        if allowlist.contains(tool_name) {
            return false;
        }

        // Default: supervised mode requires approval.
        true
    }

    /// Record an approval decision and update session state.
    pub fn record_decision(
        &self,
        tool_name: &str,
        args: &serde_json::Value,
        decision: ApprovalResponse,
        channel: &str,
    ) {
        // If "Always", add to session allowlist.
        if decision == ApprovalResponse::Always {
            let mut allowlist = self.session_allowlist.lock();
            allowlist.insert(tool_name.to_string());
        }

        // On denial, record a UserOverride correction against the tool's domain
        // so repeated rejections reduce effective autonomy for that tool.
        if decision == ApprovalResponse::No {
            if let Some(trust) = &self.trust {
                trust.lock().record_correction(
                    tool_name,
                    CorrectionType::UserOverride,
                    &format!("approval denied via {channel}"),
                );
            }
        }

        // Append to audit log.
        let summary = summarize_args(args);
        let entry = ApprovalLogEntry {
            timestamp: Utc::now().to_rfc3339(),
            tool_name: tool_name.to_string(),
            arguments_summary: summary,
            decision,
            channel: channel.to_string(),
        };
        let mut log = self.audit_log.lock();
        log.push(entry);
    }

    /// Get a snapshot of the audit log.
    pub fn audit_log(&self) -> Vec<ApprovalLogEntry> {
        self.audit_log.lock().clone()
    }

    /// Get the current session allowlist.
    pub fn session_allowlist(&self) -> HashSet<String> {
        self.session_allowlist.lock().clone()
    }

    /// Prompt the user on the CLI and return their decision.
    ///
    /// Only called for interactive (CLI) managers. Non-interactive managers
    /// auto-deny in the tool-call loop before reaching this point.
    pub fn prompt_cli(&self, request: &ApprovalRequest) -> ApprovalResponse {
        prompt_cli_interactive(request)
    }
}

// ── CLI prompt ───────────────────────────────────────────────────

/// Display the approval prompt and read user input from stdin.
fn prompt_cli_interactive(request: &ApprovalRequest) -> ApprovalResponse {
    let summary = summarize_args(&request.arguments);
    eprintln!();
    eprintln!("🔧 Agent wants to execute: {}", request.tool_name);
    eprintln!("   {summary}");
    eprint!("   [Y]es / [N]o / [A]lways for {}: ", request.tool_name);
    let _ = io::stderr().flush();

    let stdin = io::stdin();
    let mut line = String::new();
    if stdin.lock().read_line(&mut line).is_err() {
        return ApprovalResponse::No;
    }

    match line.trim().to_ascii_lowercase().as_str() {
        "y" | "yes" => ApprovalResponse::Yes,
        "a" | "always" => ApprovalResponse::Always,
        _ => ApprovalResponse::No,
    }
}

/// Produce a short human-readable summary of tool arguments.
fn summarize_args(args: &serde_json::Value) -> String {
    match args {
        serde_json::Value::Object(map) => {
            let parts: Vec<String> = map
                .iter()
                .map(|(k, v)| {
                    let val = match v {
                        serde_json::Value::String(s) => truncate_for_summary(s, 80),
                        other => {
                            let s = other.to_string();
                            truncate_for_summary(&s, 80)
                        }
                    };
                    format!("{k}: {val}")
                })
                .collect();
            parts.join(", ")
        }
        other => {
            let s = other.to_string();
            truncate_for_summary(&s, 120)
        }
    }
}

fn truncate_for_summary(input: &str, max_chars: usize) -> String {
    let mut chars = input.chars();
    let truncated: String = chars.by_ref().take(max_chars).collect();
    if chars.next().is_some() {
        format!("{truncated}")
    } else {
        input.to_string()
    }
}

// ── Tests ────────────────────────────────────────────────────────

#[cfg(test)]
mod tests {
    use super::*;
    use crate::config::AutonomyConfig;

    fn supervised_config() -> AutonomyConfig {
        AutonomyConfig {
            level: AutonomyLevel::Supervised,
            auto_approve: vec!["file_read".into(), "weather".into()],
            always_ask: vec!["shell".into()],
            ..AutonomyConfig::default()
        }
    }

    fn full_config() -> AutonomyConfig {
        AutonomyConfig {
            level: AutonomyLevel::Full,
            ..AutonomyConfig::default()
        }
    }

    // ── needs_approval ───────────────────────────────────────

    #[test]
    fn auto_approve_tools_skip_prompt() {
        let mgr = ApprovalManager::from_config(&supervised_config());
        assert!(!mgr.needs_approval("file_read"));
        assert!(!mgr.needs_approval("weather"));
    }

    #[test]
    fn always_ask_tools_always_prompt() {
        let mgr = ApprovalManager::from_config(&supervised_config());
        assert!(mgr.needs_approval("shell"));
    }

    #[test]
    fn unknown_tool_needs_approval_in_supervised() {
        let mgr = ApprovalManager::from_config(&supervised_config());
        assert!(mgr.needs_approval("file_write"));
        assert!(mgr.needs_approval("http_request"));
    }

    #[test]
    fn full_autonomy_never_prompts() {
        let mgr = ApprovalManager::from_config(&full_config());
        assert!(!mgr.needs_approval("shell"));
        assert!(!mgr.needs_approval("file_write"));
        assert!(!mgr.needs_approval("anything"));
    }

    #[test]
    fn readonly_never_prompts() {
        let config = AutonomyConfig {
            level: AutonomyLevel::ReadOnly,
            ..AutonomyConfig::default()
        };
        let mgr = ApprovalManager::from_config(&config);
        assert!(!mgr.needs_approval("shell"));
    }

    // ── session allowlist ────────────────────────────────────

    #[test]
    fn always_response_adds_to_session_allowlist() {
        let mgr = ApprovalManager::from_config(&supervised_config());
        assert!(mgr.needs_approval("file_write"));

        mgr.record_decision(
            "file_write",
            &serde_json::json!({"path": "test.txt"}),
            ApprovalResponse::Always,
            "cli",
        );

        // Now file_write should be in session allowlist.
        assert!(!mgr.needs_approval("file_write"));
    }

    #[test]
    fn always_ask_overrides_session_allowlist() {
        let mgr = ApprovalManager::from_config(&supervised_config());

        // Even after "Always" for shell, it should still prompt.
        mgr.record_decision(
            "shell",
            &serde_json::json!({"command": "ls"}),
            ApprovalResponse::Always,
            "cli",
        );

        // shell is in always_ask, so it still needs approval.
        assert!(mgr.needs_approval("shell"));
    }

    #[test]
    fn yes_response_does_not_add_to_allowlist() {
        let mgr = ApprovalManager::from_config(&supervised_config());
        mgr.record_decision(
            "file_write",
            &serde_json::json!({}),
            ApprovalResponse::Yes,
            "cli",
        );
        assert!(mgr.needs_approval("file_write"));
    }

    // ── audit log ────────────────────────────────────────────

    // ── trust integration ────────────────────────────────────

    #[test]
    fn trust_regression_downgrades_full_to_supervised_approval() {
        use crate::trust::{CorrectionType, TrustConfig, TrustTracker};

        let config = AutonomyConfig {
            level: AutonomyLevel::Full,
            ..AutonomyConfig::default()
        };
        let trust = Arc::new(Mutex::new(TrustTracker::new(TrustConfig::default())));
        let mgr = ApprovalManager::from_config(&config).with_trust_tracker(Arc::clone(&trust));

        // With no corrections, Full autonomy skips approval for any tool.
        assert!(!mgr.needs_approval("shell"));

        // Drive the tool's domain into regression.
        for _ in 0..10 {
            trust.lock().record_correction(
                "shell",
                CorrectionType::UserOverride,
                "test regression",
            );
        }

        // Regression downgrades Full → Supervised, so unknown tool now prompts.
        assert!(mgr.needs_approval("shell"));
        // Other, unaffected tool still benefits from Full autonomy.
        assert!(!mgr.needs_approval("file_write"));
    }

    #[test]
    fn trust_regression_in_supervised_forces_prompt_not_silent_bypass() {
        // Regression: before this fix, seven denials in Supervised mode
        // downgraded the tool's per-tool level to ReadOnly. The ReadOnly
        // branch in needs_approval returned false under the old comment
        // "ReadOnly blocks everything — handled elsewhere", but "elsewhere"
        // (SecurityPolicy::can_act) only inspects GLOBAL autonomy. The
        // denied tool then executed without a prompt — the exact inverse of
        // the feature's intent.
        use crate::trust::{CorrectionType, TrustConfig, TrustTracker};

        let trust = Arc::new(Mutex::new(TrustTracker::new(TrustConfig::default())));
        let mgr = ApprovalManager::from_config(&supervised_config())
            .with_trust_tracker(Arc::clone(&trust));

        // Drive the tool's domain into regression. Default config:
        // initial_score=0.8, correction_penalty=0.05, threshold=0.5 →
        // 7 denials put us at 0.45, below threshold.
        for _ in 0..7 {
            trust
                .lock()
                .record_correction("file_write", CorrectionType::UserOverride, "test");
        }

        assert!(
            mgr.needs_approval("file_write"),
            "per-tool regression (Supervised→ReadOnly) must still prompt"
        );
        // Sanity: global ReadOnly still returns false (blocked downstream).
        let readonly_config = AutonomyConfig {
            level: AutonomyLevel::ReadOnly,
            ..AutonomyConfig::default()
        };
        let readonly_mgr =
            ApprovalManager::from_config(&readonly_config).with_trust_tracker(Arc::clone(&trust));
        assert!(!readonly_mgr.needs_approval("file_write"));
    }

    #[test]
    fn denial_records_user_override_correction_when_trust_attached() {
        use crate::trust::{TrustConfig, TrustTracker};

        let trust = Arc::new(Mutex::new(TrustTracker::new(TrustConfig::default())));
        let mgr = ApprovalManager::from_config(&supervised_config())
            .with_trust_tracker(Arc::clone(&trust));

        let before = trust.lock().get_score("file_write");
        mgr.record_decision(
            "file_write",
            &serde_json::json!({"path": "x"}),
            ApprovalResponse::No,
            "cli",
        );
        let after = trust.lock().get_score("file_write");
        assert!(after < before, "denial should reduce trust score");
    }

    #[test]
    fn audit_log_records_decisions() {
        let mgr = ApprovalManager::from_config(&supervised_config());

        mgr.record_decision(
            "shell",
            &serde_json::json!({"command": "rm -rf ./build/"}),
            ApprovalResponse::No,
            "cli",
        );
        mgr.record_decision(
            "file_write",
            &serde_json::json!({"path": "out.txt", "content": "hello"}),
            ApprovalResponse::Yes,
            "cli",
        );

        let log = mgr.audit_log();
        assert_eq!(log.len(), 2);
        assert_eq!(log[0].tool_name, "shell");
        assert_eq!(log[0].decision, ApprovalResponse::No);
        assert_eq!(log[1].tool_name, "file_write");
        assert_eq!(log[1].decision, ApprovalResponse::Yes);
    }

    #[test]
    fn audit_log_contains_timestamp_and_channel() {
        let mgr = ApprovalManager::from_config(&supervised_config());
        mgr.record_decision(
            "shell",
            &serde_json::json!({"command": "ls"}),
            ApprovalResponse::Yes,
            "telegram",
        );

        let log = mgr.audit_log();
        assert_eq!(log.len(), 1);
        assert!(!log[0].timestamp.is_empty());
        assert_eq!(log[0].channel, "telegram");
    }

    // ── summarize_args ───────────────────────────────────────

    #[test]
    fn summarize_args_object() {
        let args = serde_json::json!({"command": "ls -la", "cwd": "/tmp"});
        let summary = summarize_args(&args);
        assert!(summary.contains("command: ls -la"));
        assert!(summary.contains("cwd: /tmp"));
    }

    #[test]
    fn summarize_args_truncates_long_values() {
        let long_val = "x".repeat(200);
        let args = serde_json::json!({ "content": long_val });
        let summary = summarize_args(&args);
        assert!(summary.contains(''));
        assert!(summary.len() < 200);
    }

    #[test]
    fn summarize_args_unicode_safe_truncation() {
        let long_val = "🦀".repeat(120);
        let args = serde_json::json!({ "content": long_val });
        let summary = summarize_args(&args);
        assert!(summary.contains("content:"));
        assert!(summary.contains(''));
    }

    #[test]
    fn summarize_args_non_object() {
        let args = serde_json::json!("just a string");
        let summary = summarize_args(&args);
        assert!(summary.contains("just a string"));
    }

    // ── non-interactive (channel) mode ────────────────────────

    #[test]
    fn non_interactive_manager_reports_non_interactive() {
        let mgr = ApprovalManager::for_non_interactive(&supervised_config());
        assert!(mgr.is_non_interactive());
    }

    #[test]
    fn interactive_manager_reports_interactive() {
        let mgr = ApprovalManager::from_config(&supervised_config());
        assert!(!mgr.is_non_interactive());
    }

    #[test]
    fn non_interactive_auto_approve_tools_skip_approval() {
        let mgr = ApprovalManager::for_non_interactive(&supervised_config());
        // auto_approve tools (file_read, weather) should not need approval.
        assert!(!mgr.needs_approval("file_read"));
        assert!(!mgr.needs_approval("weather"));
    }

    #[test]
    fn non_interactive_shell_skips_outer_approval_by_default() {
        let mgr = ApprovalManager::for_non_interactive(&AutonomyConfig::default());
        assert!(!mgr.needs_approval("shell"));
    }

    #[test]
    fn non_interactive_always_ask_tools_need_approval() {
        let mgr = ApprovalManager::for_non_interactive(&supervised_config());
        // always_ask tools (shell) still report as needing approval,
        // so the tool-call loop will auto-deny them in non-interactive mode.
        assert!(mgr.needs_approval("shell"));
    }

    #[test]
    fn non_interactive_unknown_tools_need_approval_in_supervised() {
        let mgr = ApprovalManager::for_non_interactive(&supervised_config());
        // Unknown tools in supervised mode need approval (will be auto-denied
        // by the tool-call loop for non-interactive managers).
        assert!(mgr.needs_approval("file_write"));
        assert!(mgr.needs_approval("http_request"));
    }

    #[test]
    fn non_interactive_full_autonomy_never_needs_approval() {
        let mgr = ApprovalManager::for_non_interactive(&full_config());
        // Full autonomy means no approval needed, even in non-interactive mode.
        assert!(!mgr.needs_approval("shell"));
        assert!(!mgr.needs_approval("file_write"));
        assert!(!mgr.needs_approval("anything"));
    }

    #[test]
    fn non_interactive_readonly_never_needs_approval() {
        let config = AutonomyConfig {
            level: AutonomyLevel::ReadOnly,
            ..AutonomyConfig::default()
        };
        let mgr = ApprovalManager::for_non_interactive(&config);
        // ReadOnly blocks execution elsewhere; approval manager does not prompt.
        assert!(!mgr.needs_approval("shell"));
    }

    #[test]
    fn non_interactive_session_allowlist_still_works() {
        let mgr = ApprovalManager::for_non_interactive(&supervised_config());
        assert!(mgr.needs_approval("file_write"));

        // Simulate an "Always" decision (would come from a prior channel run
        // if the tool was auto-approved somehow, e.g. via config change).
        mgr.record_decision(
            "file_write",
            &serde_json::json!({"path": "test.txt"}),
            ApprovalResponse::Always,
            "telegram",
        );

        assert!(!mgr.needs_approval("file_write"));
    }

    #[test]
    fn non_interactive_always_ask_overrides_session_allowlist() {
        let mgr = ApprovalManager::for_non_interactive(&supervised_config());

        mgr.record_decision(
            "shell",
            &serde_json::json!({"command": "ls"}),
            ApprovalResponse::Always,
            "telegram",
        );

        // shell is in always_ask, so it still needs approval even after "Always".
        assert!(mgr.needs_approval("shell"));
    }

    // ── ApprovalResponse serde ───────────────────────────────

    #[test]
    fn approval_response_serde_roundtrip() {
        let json = serde_json::to_string(&ApprovalResponse::Always).unwrap();
        assert_eq!(json, "\"always\"");
        let parsed: ApprovalResponse = serde_json::from_str("\"no\"").unwrap();
        assert_eq!(parsed, ApprovalResponse::No);
    }

    // ── ApprovalRequest ──────────────────────────────────────

    #[test]
    fn approval_request_serde() {
        let req = ApprovalRequest {
            tool_name: "shell".into(),
            arguments: serde_json::json!({"command": "echo hi"}),
        };
        let json = serde_json::to_string(&req).unwrap();
        let parsed: ApprovalRequest = serde_json::from_str(&json).unwrap();
        assert_eq!(parsed.tool_name, "shell");
    }

    // ── Regression: #4247 default approved tools in channels ──

    #[test]
    fn non_interactive_allows_default_auto_approve_tools() {
        let config = AutonomyConfig::default();
        let mgr = ApprovalManager::for_non_interactive(&config);

        for tool in &config.auto_approve {
            assert!(
                !mgr.needs_approval(tool),
                "default auto_approve tool '{tool}' should not need approval in non-interactive mode"
            );
        }
    }

    #[test]
    fn non_interactive_denies_unknown_tools() {
        let config = AutonomyConfig::default();
        let mgr = ApprovalManager::for_non_interactive(&config);
        assert!(
            mgr.needs_approval("some_unknown_tool"),
            "unknown tool should need approval"
        );
    }

    #[test]
    fn non_interactive_weather_is_auto_approved() {
        let config = AutonomyConfig::default();
        let mgr = ApprovalManager::for_non_interactive(&config);
        assert!(
            !mgr.needs_approval("weather"),
            "weather tool must not need approval — it is in the default auto_approve list"
        );
    }

    #[test]
    fn always_ask_overrides_auto_approve() {
        let mut config = AutonomyConfig::default();
        config.always_ask = vec!["weather".into()];
        let mgr = ApprovalManager::for_non_interactive(&config);
        assert!(
            mgr.needs_approval("weather"),
            "always_ask must override auto_approve"
        );
    }

    // ── MCP-namespaced tools in non-interactive mode ─────────

    #[test]
    fn non_interactive_auto_approves_mcp_tools() {
        let config = AutonomyConfig::default();
        let mgr = ApprovalManager::for_non_interactive(&config);
        assert!(
            !mgr.needs_approval("kumiho-memory__kumiho_memory_engage"),
            "MCP-namespaced tools should be auto-approved in non-interactive mode"
        );
        assert!(
            !mgr.needs_approval("operator-tools__create_agent"),
            "MCP-namespaced tools should be auto-approved in non-interactive mode"
        );
    }

    #[test]
    fn interactive_mcp_tools_still_need_approval() {
        let config = supervised_config();
        let mgr = ApprovalManager::from_config(&config);
        assert!(
            mgr.needs_approval("kumiho-memory__kumiho_memory_engage"),
            "MCP tools should still need approval in interactive supervised mode"
        );
    }

    #[test]
    fn non_interactive_always_ask_overrides_mcp_auto_approve() {
        let mut config = AutonomyConfig::default();
        config.always_ask = vec!["kumiho-memory__kumiho_memory_engage".into()];
        let mgr = ApprovalManager::for_non_interactive(&config);
        assert!(
            mgr.needs_approval("kumiho-memory__kumiho_memory_engage"),
            "always_ask should override MCP auto-approve"
        );
    }
}