agentic-qe 3.10.5 → 3.10.6

package/.claude/agents/v3/qe-accessibility-auditor.md

@@ -166,6 +166,14 @@ mkdir -p docs/accessibility/captions/{page-slug}
 
 **If ANY checkbox is NO and videos were detected, the task is INCOMPLETE.**
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Run multiple accessibility tools simultaneously (axe, pa11y, Lighthouse).

package/.claude/agents/v3/qe-bdd-generator.md

@@ -38,6 +38,14 @@ Proceed with generation without confirmation when acceptance criteria are clear.
 Apply pairwise example discovery automatically for scenario outlines.
 Generate step definition stubs for missing steps by default.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Generate scenarios across multiple user stories simultaneously.

package/.claude/agents/v3/qe-chaos-engineer.md

@@ -42,6 +42,14 @@ Proceed with fault injection without confirmation when blast radius is controlle
 Apply progressive chaos (start small, increase intensity).
 Always validate steady-state before and after experiments.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Run multiple independent chaos experiments simultaneously.

package/.claude/agents/v3/qe-code-complexity.md

@@ -38,6 +38,14 @@ Proceed with hotspot detection without confirmation when thresholds are configur
 Apply trend analysis automatically for repositories with history.
 Generate refactoring suggestions by default for high-complexity functions.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze complexity across multiple files simultaneously.

package/.claude/agents/v3/qe-code-intelligence.md

@@ -39,6 +39,14 @@ Proceed with analysis without confirmation when scope is clear.
 Apply incremental indexing for known codebases automatically.
 Use HNSW indexing for all semantic operations (5,900x faster at scale).
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Parse multiple source files simultaneously using worker pool.

package/.claude/agents/v3/qe-contract-validator.md

@@ -42,6 +42,14 @@ Proceed with verification without confirmation when contracts are clear.
 Apply consumer-driven principles (consumers define expectations).
 Use strict mode for production contracts, relaxed for development.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Verify multiple provider contracts simultaneously.

package/.claude/agents/v3/qe-coverage-specialist.md

@@ -61,6 +61,14 @@ Proceed with analysis without asking for confirmation when targets are specified
 Apply sublinear algorithms automatically for large codebases (>1000 files).
 Use HNSW indexing for all similarity-based operations.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze multiple directories simultaneously using worker pool.

package/.claude/agents/v3/qe-defect-predictor.md

@@ -39,6 +39,14 @@ Proceed with prediction without confirmation when context is clear.
 Apply ensemble models automatically for higher confidence.
 Use historical data to calibrate predictions continuously.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze multiple files for defect probability simultaneously.

package/.claude/agents/v3/qe-dependency-mapper.md

@@ -38,6 +38,14 @@ Proceed with analysis without confirmation when targets are clear.
 Apply circular dependency detection automatically.
 Check external dependencies for security vulnerabilities by default.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze dependencies across multiple modules simultaneously.

package/.claude/agents/v3/qe-deployment-advisor.md

@@ -52,6 +52,14 @@ Proceed with assessment without confirmation when policies are configured.
 Apply rollback planning automatically for production deployments.
 Use multi-source risk aggregation by default for comprehensive assessment.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Evaluate multiple quality gates simultaneously.

package/.claude/agents/v3/qe-devils-advocate.md

@@ -43,6 +43,14 @@ Filter results by configured minConfidence and minSeverity.
 Report challenges in descending severity order.
 Always produce a summary even when no challenges are found.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Run all applicable challenge strategies concurrently against the target.

package/.claude/agents/v3/qe-flaky-hunter.md

@@ -44,6 +44,14 @@ Proceed with remediation without confirmation for known patterns.
 Apply auto-fixes automatically for confident pattern matches.
 Use quarantine as last resort (prefer fixing over isolation).
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze multiple test suites for flakiness simultaneously.

package/.claude/agents/v3/qe-fleet-commander.md

@@ -57,6 +57,14 @@ Proceed with workload rebalancing without confirmation when thresholds are excee
 Apply autoscaling rules automatically when configured.
 Generate fleet reports by default on significant state changes.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Monitor all domain clusters simultaneously.

package/.claude/agents/v3/qe-gap-detector.md

@@ -46,6 +46,14 @@ Proceed with analysis without confirmation when scope is clear.
 Apply semantic gap detection for error handling and edge cases automatically.
 Generate test recommendations with effort estimates by default.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze coverage gaps across multiple files simultaneously.

package/.claude/agents/v3/qe-graphql-tester.md

@@ -38,6 +38,14 @@ Proceed with security scanning without confirmation for all GraphQL endpoints.
 Apply complexity analysis automatically for production queries.
 Generate test suggestions by default for uncovered fields.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Test multiple queries and mutations simultaneously.

package/.claude/agents/v3/qe-impact-analyzer.md

@@ -49,6 +49,14 @@ Proceed with test selection without confirmation when strategy is configured.
 Apply risk assessment automatically for all impact analyses.
 Generate CI/CD recommendations by default.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze impact across multiple change paths simultaneously.

package/.claude/agents/v3/qe-integration-tester.md

@@ -39,6 +39,14 @@ Proceed with test generation without confirmation when boundaries are clear.
 Apply minimal mocking strategy (prefer real dependencies where feasible).
 Use database-per-test isolation for data integrity.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Generate integration tests for independent components simultaneously.

package/.claude/agents/v3/qe-kg-builder.md

@@ -38,6 +38,14 @@ Proceed with construction without confirmation when scope is clear.
 Apply incremental updates for known repositories automatically.
 Use HNSW indexing for all graph queries by default.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Parse source files across multiple languages simultaneously.

package/.claude/agents/v3/qe-learning-coordinator.md

@@ -39,6 +39,14 @@ Proceed with knowledge synthesis without confirmation when patterns are clear.
 Apply federated aggregation for multi-agent learnings automatically.
 Use HNSW indexing for all pattern storage and retrieval.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Process learning data from multiple agents simultaneously.

package/.claude/agents/v3/qe-load-tester.md

@@ -38,6 +38,14 @@ Proceed with baseline establishment without confirmation when metrics are availa
 Apply performance assertions automatically based on SLA requirements.
 Generate bottleneck analysis by default after test completion.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Execute load tests across multiple endpoints simultaneously.

package/.claude/agents/v3/qe-message-broker-tester.md

@@ -46,6 +46,14 @@ Test DLQ routing automatically when dead letter configuration is detected.
 Validate message ordering for all FIFO-enabled queues.
 Use transactional sends by default for reliability testing.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Test multiple queues and topics simultaneously across brokers.

package/.claude/agents/v3/qe-metrics-optimizer.md

@@ -38,6 +38,14 @@ Proceed with A/B testing without confirmation when hypotheses are defined.
 Apply anomaly detection automatically for all monitored agents.
 Generate optimization recommendations by default after analysis.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze multiple agents simultaneously.

package/.claude/agents/v3/qe-middleware-validator.md

@@ -44,6 +44,14 @@ Test error handling paths automatically for every flow branch.
 Validate protocol mediation mappings between source and target formats.
 Use service virtualization to isolate middleware from backend dependencies during testing.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Test multiple message flows simultaneously across ESB nodes.

package/.claude/agents/v3/qe-mutation-tester.md

@@ -38,6 +38,14 @@ Proceed with surviving mutant analysis without confirmation after test completio
 Apply representative sampling automatically for large codebases.
 Generate test improvement suggestions by default for weak tests.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Execute mutation tests across multiple mutants simultaneously.

package/.claude/agents/v3/qe-odata-contract-tester.md

@@ -48,6 +48,14 @@ Automatically fetch CSRF tokens before mutation operations on SAP services.
 Generate test cases from $metadata document without manual entity specification.
 Detect SAP Gateway vs standard OData services and apply appropriate extensions.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Validate multiple entity sets simultaneously from a single $metadata document.

package/.claude/agents/v3/qe-parallel-executor.md

@@ -39,6 +39,14 @@ Proceed with execution without confirmation when test targets are clear.
 Apply time-balanced sharding automatically for optimal distribution.
 Use dynamic rebalancing to handle slow tests.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Execute tests across multiple workers simultaneously (up to 16).

package/.claude/agents/v3/qe-pattern-learner.md

@@ -38,6 +38,14 @@ Proceed with learning without confirmation when confidence thresholds are met.
 Apply incremental updates automatically as new data arrives.
 Use ensemble methods by default for robust pattern detection.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Process pattern discovery across multiple domains simultaneously.

package/.claude/agents/v3/qe-pentest-validator.md

@@ -68,6 +68,14 @@ Never report a vulnerability without exploitation evidence.
 Require explicit target authorization before any exploitation.
 Sandbox enforcement: only test against declared staging/dev URLs.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Run per-vulnerability-type pipelines in parallel:

package/.claude/agents/v3/qe-performance-tester.md

@@ -39,6 +39,14 @@ Proceed with testing without confirmation when thresholds are clear.
 Apply statistical analysis to all benchmark results automatically.
 Use multi-scenario testing by default for comprehensive coverage.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Execute multiple performance scenarios simultaneously.

package/.claude/agents/v3/qe-product-factors-assessor.md

@@ -41,6 +41,14 @@ Use domain-specific patterns for test idea generation.
 Always read HTML template before generating HTML output.
 Output complete assessments in requested format.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze all 7 SFDIPOT categories simultaneously.

package/.claude/agents/v3/qe-property-tester.md

@@ -38,6 +38,14 @@ Proceed with testing without confirmation when properties are clear.
 Apply shrinking automatically to all counterexamples.
 Use multiple runs (100+ iterations) by default for statistical confidence.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Execute property tests across multiple arbitraries simultaneously.

package/.claude/agents/v3/qe-quality-criteria-recommender.md

@@ -42,6 +42,14 @@ Collect evidence with file:line references.
 Quantify business impact with citations.
 Always read HTML template before generating HTML output.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze multiple HTSM categories simultaneously.

package/.claude/agents/v3/qe-quality-gate.md

@@ -39,6 +39,14 @@ Proceed with gate evaluation without confirmation when thresholds are clear.
 Apply learned patterns for risk assessment automatically.
 Use strict mode by default, allow overrides with proper approval.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Evaluate multiple gate criteria simultaneously.

package/.claude/agents/v3/qe-qx-partner.md

@@ -111,6 +111,14 @@ Generate QX recommendations by default for all significant quality events.
 **ALWAYS persist patterns** - save JSON to .agentic-qe/qx-patterns/ for cross-session learning.
 **ALWAYS use fetch-content.js cascade for URL analysis** - never manually retry failed browser operations.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze multiple user journeys simultaneously.

package/.claude/agents/v3/qe-regression-analyzer.md

@@ -38,6 +38,14 @@ Proceed with historical analysis without confirmation when data is available.
 Apply risk scoring automatically for all change sets.
 Generate test recommendations by default with execution time estimates.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze multiple change sets simultaneously.

package/.claude/agents/v3/qe-requirements-validator.md

@@ -43,6 +43,14 @@ Apply vague term detection automatically for all requirements.
 Generate traceability reports by default for test-linked requirements.
 Block requirements scoring below 50/100 from proceeding to development.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Analyze multiple requirements simultaneously.

package/.claude/agents/v3/qe-responsive-tester.md

@@ -38,6 +38,14 @@ Proceed with breakpoint validation without confirmation when breakpoints are def
 Apply touch target validation automatically for mobile viewports.
 Generate responsive reports by default with screenshots at each breakpoint.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Test multiple viewports simultaneously.

package/.claude/agents/v3/qe-retry-handler.md

@@ -38,6 +38,14 @@ Proceed with retries without confirmation when patterns match known transient is
 Apply circuit breakers automatically for cascading failures.
 Use adaptive backoff by default for resource-related failures.
 </default_to_action>
+<evidence_discipline>
+ADR-105 evidence classes — label every finding you emit:
+- EXECUTED: you ran a real command; attach the command and its output as the artifact.
+- STATIC: derived from data (coverage file, AST, lockfile, schema); name the data source.
+- INFERRED: reasoning over code/content without execution. Never present it in the voice of verified fact.
+- CONJECTURE: pattern-matched heuristic or extrapolation; flag it as such.
+Quality gates block only on EXECUTED/STATIC; INFERRED routes to adversarial verification (ADR-102); CONJECTURE never gates. When a check can cheaply be executed instead of inferred, execute it and upgrade the label.
+</evidence_discipline>
 
 <parallel_execution>
 Execute retries across multiple tests simultaneously.