principles-disciple 1.104.0 → 1.104.2

package/src/hooks/triage-adapter.ts

@@ -28,81 +28,54 @@ import {
 // ── Source Kind Resolution ───────────────────────────────────────────────────
 
 /**
- * Map after_tool_call hook context to SourceKind.
+ * Map RawObservation to SourceKind.
  *
- * Classifies based on:
- * - toolName: 'pain' or 'skill:pain' → agent_on_owner_request
- * - failureSource: 'dispatch_error' vs 'tool_failure'
- * - isRisky + score: only used for rulehost_block upgrade, not for kind resolution
+ * This is the unified entry point for source-kind classification.
+ * It replaces the scattered resolveSourceKindFrom* functions.
  */
-export function resolveSourceKindFromToolFailure(
-  toolName: string | undefined,
-  failureSource: 'tool_failure' | 'dispatch_error',
-  provenance?: 'openclaw_context_bound' | 'owner_reported_no_host_trace' | 'automatic_hook',
-): SourceKind {
-  // Manual pain via agent tool call
-  if (toolName === 'pain' || toolName === 'skill:pain') {
-    return provenance === 'openclaw_context_bound' ? 'agent_on_owner_request' : 'owner_reported';
-  }
-
-  // Dispatch errors (tool not found, unknown tool)
-  if (failureSource === 'dispatch_error') {
-    return 'dispatch_error';
-  }
+export { resolveSourceKind } from './raw-observation-adapter.js';
 
-  // Regular tool failure
-  return 'tool_failure';
-}
+/**
+ * Map after_tool_call hook context to SourceKind.
+ *
+ * @deprecated Use resolveSourceKind directly with RawObservation.
+ */
+export { resolveSourceKindFromToolFailure } from './raw-observation-adapter.js';
 
 /**
  * Map empathy/semantic detection context to SourceKind.
  *
- * Classifies based on detection source prefix:
- * - 'llm_paralysis' → llm_paralysis
- * - 'llm_*' (detection rule) → semantic
- * - 'user_empathy' or empathy keyword match → empathy_inferred
- * - GFI threshold crossed → gfi_threshold
+ * @deprecated Use resolveSourceKind directly with RawObservation.
  */
-export function resolveSourceKindFromLlmDetection(
-  detectionSource: string,
-  isGfiTriggered: boolean,
-): SourceKind {
-  if (isGfiTriggered) return 'gfi_threshold';
-  if (detectionSource === 'llm_paralysis') return 'llm_paralysis';
-  if (detectionSource.startsWith('llm_')) return 'semantic';
-  if (detectionSource === 'user_empathy') return 'empathy_inferred';
-  return 'unknown';
-}
+export { resolveSourceKindFromLlmDetection } from './raw-observation-adapter.js';
 
 /**
  * Map gate-block context to SourceKind.
+ *
+ * @deprecated Use resolveSourceKind directly with RawObservation.
  */
-export function resolveSourceKindFromGateBlock(): SourceKind {
-  return 'rulehost_block';
-}
+export { resolveSourceKindFromGateBlock } from './raw-observation-adapter.js';
 
 /**
  * Map /pd-pain command to SourceKind.
+ *
+ * @deprecated Use resolveSourceKind directly with RawObservation.
  */
-export function resolveSourceKindFromCommand(): SourceKind {
-  return 'owner_reported';
-}
+export { resolveSourceKindFromCommand } from './raw-observation-adapter.js';
 
 /**
  * Map provider/rate-limit failure to SourceKind.
+ *
+ * @deprecated Use resolveSourceKind directly with RawObservation.
  */
-export function resolveSourceKindFromProvider(
-  isRateLimit: boolean,
-): SourceKind {
-  return isRateLimit ? 'rate_limit' : 'provider_failure';
-}
+export { resolveSourceKindFromProvider } from './raw-observation-adapter.js';
 
 /**
  * Map subagent error to SourceKind.
+ *
+ * @deprecated Use resolveSourceKind directly with RawObservation.
  */
-export function resolveSourceKindFromSubagent(): SourceKind {
-  return 'subagent_error';
-}
+export { resolveSourceKindFromSubagent } from './raw-observation-adapter.js';
 
 // ── Triage Evaluation ───────────────────────────────────────────────────────
 
@@ -123,6 +96,8 @@ export function evaluateEvidenceTriage(
   options?: {
     isUnsafeHighConfidence?: boolean;
     provenance?: 'openclaw_context_bound' | 'owner_reported_no_host_trace' | 'automatic_hook';
+    consecutiveErrors?: number;
+    isRisky?: boolean;
   },
 ): TriageResult {
   const input: TriageInput = {
@@ -132,7 +107,39 @@ export function evaluateEvidenceTriage(
     provenance: options?.provenance,
   };
 
-  return evaluateTriage(input);
+  let result = evaluateTriage(input);
+
+  // PEAT-B1 upgrade logic: risky high-score overrides evidence_only
+  // Matches PainDiagnosticGate.risky_high_score: isRisky && score >= 70 → admit
+  if (
+    result.decision === 'evidence_only' &&
+    options?.isRisky === true &&
+    score >= 70
+  ) {
+    result = {
+      ...result,
+      decision: 'admit',
+      reason: 'Risky high-score operation overrides evidence-only decision. Immediate diagnosis required.',
+      nextAction: 'create_diagnostic_task',
+    };
+  }
+
+  // PEAT-B1 upgrade logic: repeated failures override evidence_only
+  // Threshold: 4 consecutive failures (matches PainDiagnosticGate.repeatedFailure)
+  if (
+    result.decision === 'evidence_only' &&
+    options?.consecutiveErrors !== undefined &&
+    options.consecutiveErrors >= 4
+  ) {
+    result = {
+      ...result,
+      decision: 'admit',
+      reason: 'Repeated failures override evidence-only decision. Pattern suggests systemic issue requiring diagnosis.',
+      nextAction: 'create_diagnostic_task',
+    };
+  }
+
+  return result;
 }
 
 // ── High-Confidence Unsafe Action Detection ──────────────────────────────────

package/src/hooks/trigger-cooldown-tracker.ts

@@ -0,0 +1,82 @@
+/**
+ * Trigger Cooldown Tracker — PRI-363
+ *
+ * Manages cooldown state for trigger controller decisions.
+ *
+ * This is a plugin-layer concern because:
+ * - Core (trigger-controller) is stateless and pure
+ * - Cooldown state needs to persist across tool calls
+ * - The map is scoped to the plugin's lifecycle
+ *
+ * EP-05: Loop State Freshness — each check reads fresh state from the map.
+ * ERR-001: No `as` casts on map access.
+ * ERR-002: Every rejected decision includes reason + nextAction.
+ */
+
+const DEFAULT_COOLDOWN_MS = 15 * 60 * 1000; // 15 minutes
+
+/**
+ * Episode key format: sessionId:source:errorHash
+ */
+function buildEpisodeKey(
+  source: string,
+  sessionId: string | undefined,
+  errorHash: string | undefined,
+): string {
+  const sid = sessionId || 'unknown';
+  const hash = errorHash || 'no-hash';
+  return `${sid}:${source}:${hash}`;
+}
+
+/**
+ * Check whether cooldown is currently active for a given episode.
+ */
+export function isCooldownActive(
+  source: string,
+  sessionId: string | undefined,
+  errorHash: string | undefined,
+  cooldownMap: ReadonlyMap<string, number>,
+): boolean {
+  const episodeKey = buildEpisodeKey(source, sessionId, errorHash);
+  const lastDiagnosedAt = cooldownMap.get(episodeKey);
+
+  if (lastDiagnosedAt === undefined) {
+    return false;
+  }
+
+  const now = Date.now();
+  return now - lastDiagnosedAt < DEFAULT_COOLDOWN_MS;
+}
+
+/**
+ * Mark an episode as diagnosed (set cooldown timestamp).
+ */
+export function markEpisodeAsDiagnosed(
+  source: string,
+  sessionId: string | undefined,
+  errorHash: string | undefined,
+  cooldownMap: Map<string, number>,
+): void {
+  const episodeKey = buildEpisodeKey(source, sessionId, errorHash);
+  cooldownMap.set(episodeKey, Date.now());
+}
+
+/**
+ * Clear all cooldown state (for tests).
+ */
+export function clearCooldownState(cooldownMap: Map<string, number>): void {
+  cooldownMap.clear();
+}
+
+/**
+ * Get the cooldown timestamp for a given episode (for tests).
+ */
+export function getCooldownTimestamp(
+  source: string,
+  sessionId: string | undefined,
+  errorHash: string | undefined,
+  cooldownMap: ReadonlyMap<string, number>,
+): number | undefined {
+  const episodeKey = buildEpisodeKey(source, sessionId, errorHash);
+  return cooldownMap.get(episodeKey);
+}

package/templates/langs/en/principles/THINKING_OS.md

@@ -62,4 +62,16 @@ LLMs are highly sensitive to XML tags; this structure is designed to boost instr
     <must>Maintain extreme digital cleanliness. The project root is SACRED. Use strict `kebab-case` for all naming. Clean up all test scripts and debug artifacts after the task.</must>
     <forbidden>Creating arbitrary temporary files (e.g., `test.txt`, `temp.md`, `debug.log`) in the project root directory.</forbidden>
   </directive>
+
+  <directive id="T-09" name="DIVIDE_AND_CONQUER">
+    <trigger>When facing a complex task, multi-step change, or an operation that can be decomposed.</trigger>
+    <must>Break the task into smaller, manageable phases before execution. Execute one phase at a time and verify each phase's result before proceeding.</must>
+    <forbidden>Attempting to execute a large, complex change in a single step, or proceeding without a decomposition plan.</forbidden>
+  </directive>
+
+  <directive id="T-10" name="MEMORY_EXTERNALIZATION">
+    <trigger>When reaching a significant conclusion, making a decision, or planning next steps across sessions.</trigger>
+    <must>Write intermediate conclusions, decisions, and plans to persistent files (e.g., plan.md, scratchpad) so they survive context compression and session boundaries.</must>
+    <forbidden>Relying solely on conversation context to retain important state that will be lost when the context window shifts.</forbidden>
+  </directive>
 </thinking_os_core_directives>

package/templates/langs/zh/principles/THINKING_OS.md

@@ -62,4 +62,16 @@
     <must>保持极致的数字洁癖。项目根目录是神圣的。所有命名必须严格使用 `kebab-case`。任务结束后清理所有的测试脚本和 Debug 遗留物。</must>
     <forbidden>在项目根目录下随意创建临时文件（如 `test.txt`、`temp.md`、`debug.log`）。</forbidden>
   </directive>
+
+  <directive id="T-09" name="DIVIDE_AND_CONQUER">
+    <trigger>当面对复杂任务、多步骤变更或可分解的操作时。</trigger>
+    <must>在执行前将任务拆分为更小的、可管理的阶段。逐阶段执行，并在进入下一阶段前验证当前阶段的结果。</must>
+    <forbidden>试图在单一步骤中执行大型复杂变更，或在没有分解计划的情况下直接推进。</forbidden>
+  </directive>
+
+  <directive id="T-10" name="MEMORY_EXTERNALIZATION">
+    <trigger>当得出重要结论、做出决策或规划跨会话的后续步骤时。</trigger>
+    <must>将中间结论、决策和计划写入持久化文件（如 plan.md、scratchpad），使其在上下文压缩和会话边界后仍然可用。</must>
+    <forbidden>仅依赖对话上下文来保持重要状态，这些状态在上下文窗口切换时将丢失。</forbidden>
+  </directive>
 </thinking_os_core_directives>

package/tests/hooks/pain.test.ts

@@ -7,7 +7,7 @@ import * as ioUtils from '../../src/utils/io.js';
 import { WorkspaceContext } from '../../src/core/workspace-context.js';
 import { EventLogService } from '../../src/core/event-log.js';
 import { setInjectedProbationIds, clearSession } from '../../src/core/session-tracker.js';
-import { resetPainDiagnosticGateForTest } from '../../src/core/pain-diagnostic-gate.js';
+import { resetTriggerCooldownForTest } from '../../src/hooks/after-tool-call-helpers.js';
 import { loadFeatureFlagFromConfig } from '../../src/core/pd-config-loader.js';
 
 vi.mock('fs');
@@ -140,12 +140,12 @@ describe('Post-Write Checks & Pain Hook', () => {
     mockEmitSync.mockReset();
     mockRecordProbationFeedback.mockReset();
     mockUpdatePrincipleValueMetrics.mockReset();
-    vi.spyOn(WorkspaceContext, 'fromHookContext').mockReturnValue(mockWctx as any);
+    vi.spyOn(WorkspaceContext, 'fromHookContextExplicit').mockReturnValue(mockWctx as any);
     vi.spyOn(EventLogService, 'get').mockReturnValue(mockEventLog as any);
     clearSession('s-success');
     clearSession('s-low-value-failure');
     clearSession('s-repeated-failure');
-    resetPainDiagnosticGateForTest();
+    resetTriggerCooldownForTest();
   });
 
   afterEach(() => {
@@ -158,7 +158,7 @@ describe('Post-Write Checks & Pain Hook', () => {
     handleAfterToolCall(mockEvent as any, mockCtx as any);
     
     // Should still create context
-    expect(WorkspaceContext.fromHookContext).toHaveBeenCalled();
+    expect(WorkspaceContext.fromHookContextExplicit).toHaveBeenCalled();
     expect(fs.writeFileSync).not.toHaveBeenCalled();
     expect(mockEmitSync).not.toHaveBeenCalled();
   });
@@ -178,7 +178,7 @@ describe('Post-Write Checks & Pain Hook', () => {
 
     handleAfterToolCall(mockEvent as any, mockCtx as any, mockApi as any);
 
-    expect(WorkspaceContext.fromHookContext).not.toHaveBeenCalled();
+    expect(WorkspaceContext.fromHookContextExplicit).not.toHaveBeenCalled();
     expect(mockEmitSync).not.toHaveBeenCalled();
   });
 
@@ -221,6 +221,13 @@ describe('Post-Write Checks & Pain Hook', () => {
     vi.mocked(ioUtils.isRisky).mockReturnValue(false);
     vi.mocked(fs.existsSync).mockReturnValue(false);
 
+    // PRI-363: trigger controller requires consecutiveErrors >= 4 for upgrade
+    handleAfterToolCall(mockEvent as any, mockCtx as any);
+    expect(mockEmitSync).not.toHaveBeenCalled();
+
+    handleAfterToolCall(mockEvent as any, mockCtx as any);
+    expect(mockEmitSync).not.toHaveBeenCalled();
+
     handleAfterToolCall(mockEvent as any, mockCtx as any);
     expect(mockEmitSync).not.toHaveBeenCalled();
 
@@ -231,7 +238,6 @@ describe('Post-Write Checks & Pain Hook', () => {
       data: expect.objectContaining({
         painType: 'tool_failure',
         source: 'write',
-        reason: expect.stringContaining('diagnosticGate=high_gfi'),
       }),
     }));
     expect(mockWctx.trajectory.recordToolCall).toHaveBeenCalledWith(expect.objectContaining({
@@ -309,6 +315,8 @@ describe('Post-Write Checks & Pain Hook', () => {
       contextTags: ['write'],
     });
 
+    // PRI-363: risky high-score write triggers admission via trigger controller
+    // (isRisky=true + score >= 70 → risky_high_score upgrade → admit)
     handleAfterToolCall(mockEvent as any, mockCtx as any);
 
     expect(mockUpdatePrincipleValueMetrics).toHaveBeenCalledWith(
@@ -652,7 +660,7 @@ describe('PRI-326: evaluatePainAdmissionForToolCall', () => {
 
   beforeEach(() => {
     vi.clearAllMocks();
-    resetPainDiagnosticGateForTest();
+    resetTriggerCooldownForTest();
     vi.mocked(loadFeatureFlagFromConfig).mockReturnValue({ enabled: false, source: 'test' });
   });
 
@@ -672,29 +680,27 @@ describe('PRI-326: evaluatePainAdmissionForToolCall', () => {
     expect(result.stage).toBe('not_applicable');
   });
 
-  it('returns triage_evidence_only when feature flag on and tool_failure triage rejects', () => {
+  it('returns trigger_rejected when tool_failure triage rejects', () => {
     vi.mocked(loadFeatureFlagFromConfig).mockReturnValue({ enabled: true, source: 'test' });
 
     const result = evaluatePainAdmissionForToolCall(
       { toolName: 'write' } as any, baseObservation, baseOutcome, undefined, undefined, 's1', workspaceDir, mockConfig
     );
-    expect(result.stage).toBe('triage_evidence_only');
+    expect(result.stage).toBe('trigger_rejected');
     expect(result.admitted).toBe(false);
     expect(result.reason).toBeTruthy();
   });
 
-  it('returns gate_admitted when consecutive errors exceed repeatedFailure threshold', () => {
+  it('returns trigger_admitted when consecutive errors exceed repeatedFailure threshold', () => {
     vi.mocked(loadFeatureFlagFromConfig).mockReturnValue({ enabled: false, source: 'test' });
-    // consecutiveErrors=5 >= default repeatedFailure threshold of 4 → gate admits via repeated_failure
+    // consecutiveErrors=5 >= default repeatedFailure threshold of 4 → trigger admits via repeated_failure
     const highConsecutiveState = { currentGfi: 0, consecutiveErrors: 5, lastErrorHash: 'abc123' } as any;
 
     const result = evaluatePainAdmissionForToolCall(
       { toolName: 'write' } as any, baseObservation, baseOutcome, highConsecutiveState, undefined, 's-gate-admitted-test', workspaceDir, mockConfig
     );
-    expect(result.stage).toBe('gate_admitted');
+    expect(result.stage).toBe('trigger_admitted');
     expect(result.admitted).toBe(true);
-    expect(result.gateResult?.shouldDiagnose).toBe(true);
-    expect(result.gateResult?.reason).toBe('repeated_failure');
   });
 
   it('includes reason and detail in every decision', () => {