opencode-swarm 7.115.0 → 7.116.0

package/dist/mutation/engine.d.ts

@@ -1,6 +1,6 @@
 import { spawnSync } from 'node:child_process';
-type SpawnSyncFn = typeof spawnSync;
-export type MutationOutcome = 'killed' | 'survived' | 'timeout' | 'error' | 'equivalent' | 'skipped';
+import { resolveExecutableFromPath, runExternalTool } from '../utils/external-tool-runner.js';
+export type MutationOutcome = 'killed' | 'survived' | 'timeout' | 'cancelled' | 'error' | 'equivalent' | 'skipped';
 /**
  * Known test runner executables permitted as the first element of testCommand.
  * Validated as the basename (without extension) to support platform-specific
@@ -35,6 +35,7 @@ export interface MutationReport {
     killed: number;
     survived: number;
     timeout: number;
+    cancelled: number;
     equivalent: number;
     skipped: number;
     errors: number;
@@ -55,13 +56,36 @@ export interface MutationReport {
     timestamp: string;
 }
 export declare const MAX_MUTATIONS_PER_FUNCTION = 10;
+type LegacySpawnSyncFn = typeof spawnSync;
+export type MutationCommandResult = {
+    status: 'completed' | 'timeout' | 'cancelled' | 'spawn-error';
+    exitCode: number | null;
+    stdout: string;
+    stderr: string;
+    message?: string;
+};
+export type MutationCommandRunner = (args: {
+    executable: string;
+    args: string[];
+    cwd: string;
+    timeoutMs: number;
+    abortSignal?: AbortSignal;
+}) => Promise<MutationCommandResult>;
+export type MutationExecutionOptions = {
+    abortSignal?: AbortSignal;
+    runner?: MutationCommandRunner;
+};
+export declare const runMutationCommand: MutationCommandRunner;
 export declare const _internals: {
     executeMutation: typeof executeMutation;
     computeReport: typeof computeReport;
     executeMutationSuite: typeof executeMutationSuite;
-    spawnSync: SpawnSyncFn;
+    spawnSync: LegacySpawnSyncFn;
+    runCommand: MutationCommandRunner;
+    runExternalTool: typeof runExternalTool;
+    resolveExecutableFromPath: typeof resolveExecutableFromPath;
 };
-export declare function executeMutation(patch: MutationPatch, testCommand: string[], testFiles: string[], workingDir: string): Promise<MutationResult>;
+export declare function executeMutation(patch: MutationPatch, testCommand: string[], testFiles: string[], workingDir: string, options?: MutationExecutionOptions): Promise<MutationResult>;
 export declare function computeReport(results: MutationResult[], durationMs: number, budgetMs?: number): MutationReport;
-export declare function executeMutationSuite(patches: MutationPatch[], testCommand: string[], testFiles: string[], workingDir: string, budgetMs?: number, onProgress?: (completed: number, total: number, result: MutationResult) => void, sourceFiles?: Map<string, string>): Promise<MutationReport>;
+export declare function executeMutationSuite(patches: MutationPatch[], testCommand: string[], testFiles: string[], workingDir: string, budgetMs?: number, onProgress?: (completed: number, total: number, result: MutationResult) => void, sourceFiles?: Map<string, string>, options?: MutationExecutionOptions): Promise<MutationReport>;
 export {};

package/dist/telemetry.d.ts

@@ -1,5 +1,14 @@
 import type { DelegationCostFields } from './services/cost-accounting.js';
-export type TelemetryEvent = 'session_started' | 'session_ended' | 'agent_activated' | 'delegation_begin' | 'delegation_end' | 'task_state_changed' | 'gate_passed' | 'gate_failed' | 'gate_parse_error' | 'phase_changed' | 'budget_updated' | 'model_fallback' | 'hard_limit_hit' | 'revision_limit_hit' | 'loop_detected' | 'scope_violation' | 'qa_skip_violation' | 'heartbeat' | 'turbo_mode_changed' | 'auto_oversight_escalation' | 'environment_detected' | 'evidence_lock_acquired' | 'evidence_lock_contended' | 'evidence_lock_stale_recovered' | 'plan_ledger_cas_retry' | 'plan_md_write_failed' | 'snapshot_failed' | 'prm_pattern_detected' | 'prm_course_correction_injected' | 'prm_escalation_triggered' | 'prm_hard_stop';
+export type TelemetryEvent = 'session_started' | 'session_ended' | 'agent_activated' | 'delegation_begin' | 'delegation_end' | 'task_state_changed' | 'gate_passed' | 'gate_failed' | 'gate_parse_error' | 'reviewer_gate_decision' | 'phase_changed' | 'budget_updated' | 'model_fallback' | 'hard_limit_hit' | 'revision_limit_hit' | 'loop_detected' | 'scope_violation' | 'qa_skip_violation' | 'heartbeat' | 'turbo_mode_changed' | 'auto_oversight_escalation' | 'environment_detected' | 'evidence_lock_acquired' | 'evidence_lock_contended' | 'evidence_lock_stale_recovered' | 'plan_ledger_cas_retry' | 'plan_md_write_failed' | 'snapshot_failed' | 'prm_pattern_detected' | 'prm_course_correction_injected' | 'prm_escalation_triggered' | 'prm_hard_stop';
+/** Stable classification for how a reviewer-gate decision was established. */
+export type ReviewerGateEvidenceKind = 'genuine' | 'fallback' | 'data_quality' | 'block';
+/**
+ * Stable reason codes for reviewer-gate decision telemetry.
+ *
+ * These values are persisted in telemetry.jsonl and consumed by offline gate
+ * analytics, so additions are allowed but existing values must not be renamed.
+ */
+export type ReviewerGateReasonCode = 'lean_turbo_completed_lane' | 'standard_turbo_non_tier3' | 'durable_evidence_complete' | 'workflow_state_complete' | 'stage_b_parallel_complete' | 'no_active_sessions' | 'zero_valid_sessions' | 'restart_recovery_complete' | 'scoped_delegation_complete' | 'unscoped_delegation_complete' | 'corrupt_evidence' | 'required_gates_missing' | 'inspection_error';
 export type TelemetryListener = (event: TelemetryEvent, data: Record<string, unknown>) => void;
 /**
  * Number of emits between rotation checks. 50 keeps the overhead at one
@@ -48,6 +57,7 @@ export declare const telemetry: {
     gatePassed(sessionId: string, gate: string, taskId: string): void;
     gateParseError(taskId: string, error: Error): void;
     gateFailed(sessionId: string, gate: string, taskId: string, reason: string): void;
+    reviewerGateDecision(sessionId: string, taskId: string, blocked: boolean, reasonCode: ReviewerGateReasonCode, evidenceKind: ReviewerGateEvidenceKind): void;
     phaseChanged(sessionId: string, oldPhase: number, newPhase: number): void;
     budgetUpdated(sessionId: string, budgetPct: number, agentName: string): void;
     modelFallback(sessionId: string, agentName: string, fromModel: string, toModel: string, reason: string): void;

package/dist/tools/plugin-registration.d.ts

@@ -11,6 +11,7 @@
 import type { ToolDefinition } from '@opencode-ai/plugin/tool';
 import type { AgentDefinition } from '../agents/index.js';
 import type { PluginConfig } from '../config/index.js';
+import type { EvaluationModelDispatcher } from '../evaluation/model-dispatcher.js';
 /**
  * Construct the plugin tool object: one handler per manifest entry, with
  * `swarm_command` overridden by its dependency-injected instance.
@@ -21,4 +22,4 @@ import type { PluginConfig } from '../config/index.js';
  *
  * Knowledge tools are conditionally excluded when config.knowledge.enabled = false.
  */
-export declare function buildPluginToolObject(agents: Record<string, AgentDefinition>, config?: PluginConfig): Record<string, ToolDefinition>;
+export declare function buildPluginToolObject(agents: Record<string, AgentDefinition>, config?: PluginConfig, evaluationModelDispatcher?: EvaluationModelDispatcher): Record<string, ToolDefinition>;

package/dist/tools/sast-scan.d.ts

@@ -5,9 +5,16 @@
 import type { ToolDefinition } from '@opencode-ai/plugin/tool';
 import type { PluginConfig } from '../config';
 import type { EvidenceVerdict } from '../config/evidence-schema';
+import { isSemgrepAvailable, runSemgrep } from '../sast/semgrep';
 export interface SastScanInput {
     /** List of files to scan */
     changed_files: string[];
+    /**
+     * Restrict execution to built-in Tier A rules. This is used by isolated,
+     * reproducible evaluation runs where PATH discovery and Semgrep execution
+     * (including remote `--config=auto`) are forbidden.
+     */
+    offline_only?: boolean;
     /** Minimum severity that causes failure (default: 'medium') */
     severity_threshold?: 'low' | 'medium' | 'high' | 'critical';
     /**
@@ -94,4 +101,6 @@ export declare const sast_scan: ToolDefinition;
 export declare const _internals: {
     sastScan: typeof sastScan;
     sast_scan: typeof sast_scan;
+    isSemgrepAvailable: typeof isSemgrepAvailable;
+    runSemgrep: typeof runSemgrep;
 };

package/dist/tools/swarm-command.d.ts

@@ -1,4 +1,5 @@
 import type { AgentDefinition } from '../agents/index.js';
+import type { EvaluationModelDispatcher } from '../evaluation/model-dispatcher.js';
 import { createSwarmTool } from './create-tool.js';
-export declare function createSwarmCommandTool(agents: Record<string, AgentDefinition>): ReturnType<typeof createSwarmTool>;
+export declare function createSwarmCommandTool(agents: Record<string, AgentDefinition>, evaluationModelDispatcher?: EvaluationModelDispatcher): ReturnType<typeof createSwarmTool>;
 export declare const swarm_command: ReturnType<typeof createSwarmTool>;

package/dist/tools/update-task-status.d.ts

@@ -3,8 +3,12 @@
  * Allows agents to mark tasks as pending, in_progress, completed, or blocked.
  */
 import type { ToolContext, ToolDefinition } from '@opencode-ai/plugin/tool';
+import { readTaskEvidenceRaw } from '../gate-evidence.js';
 import { tryAcquireLock } from '../parallel/file-locks.js';
 import { loadPlan, updateTaskStatus } from '../plan/manager';
+import { hasActiveLeanTurbo, hasActiveTurboMode } from '../state';
+import { type ReviewerGateEvidenceKind, type ReviewerGateReasonCode } from '../telemetry.js';
+import { verifyLeanTurboTaskCompletion } from '../turbo/lean/task-completion';
 import { resolveWorkingDirectory } from './resolve-working-directory';
 /**
  * Internal seams for test injection.
@@ -12,10 +16,16 @@ import { resolveWorkingDirectory } from './resolve-working-directory';
  * module-scope vi.mock, which leaks across files in Bun's shared test runner.
  */
 export declare const _internals: {
-    readonly tryAcquireLock: typeof tryAcquireLock;
-    readonly updateTaskStatus: typeof updateTaskStatus;
-    readonly resolveWorkingDirectory: typeof resolveWorkingDirectory;
-    readonly loadPlan: typeof loadPlan;
+    tryAcquireLock: typeof tryAcquireLock;
+    updateTaskStatus: typeof updateTaskStatus;
+    resolveWorkingDirectory: typeof resolveWorkingDirectory;
+    loadPlan: typeof loadPlan;
+    readTaskEvidenceRaw: typeof readTaskEvidenceRaw;
+    hasActiveLeanTurbo: typeof hasActiveLeanTurbo;
+    hasActiveTurboMode: typeof hasActiveTurboMode;
+    verifyLeanTurboTaskCompletion: typeof verifyLeanTurboTaskCompletion;
+    hasPassedDurableGateEvidence: typeof hasPassedDurableGateEvidence;
+    emitReviewerGateDecision: (sessionId: string, taskId: string, blocked: boolean, reasonCode: ReviewerGateReasonCode, evidenceKind: ReviewerGateEvidenceKind) => void;
 };
 /**
  * Arguments for the update_task_status tool
@@ -58,6 +68,7 @@ export interface ReviewerGateResult {
     blocked: boolean;
     reason: string;
 }
+declare function hasPassedDurableGateEvidence(workingDirectory: string, taskId: string): boolean;
 /**
  * Check if a task has passed required QA gates using the state machine.
  * Requires the task to be in 'tests_run' or 'complete' state, which means
@@ -138,3 +149,4 @@ export declare function executeUpdateTaskStatus(args: UpdateTaskStatusArgs, fall
  * Tool definition for update_task_status
  */
 export declare const update_task_status: ToolDefinition;
+export {};

package/dist/utils/external-tool-runner.d.ts

@@ -7,9 +7,10 @@ export interface ExternalToolRunOptions {
     maxStdoutBytes: number;
     maxStderrBytes: number;
     env?: Record<string, string | undefined>;
+    abortSignal?: AbortSignal;
 }
 export interface ExternalToolRunResult {
-    status: 'completed' | 'timeout' | 'spawn-error';
+    status: 'completed' | 'timeout' | 'cancelled' | 'spawn-error';
     exitCode: number | null;
     stdout: string;
     stderr: string;

package/evaluation-fixtures/tier1/boundary-error/environment/baseline.ts

@@ -0,0 +1,3 @@
+export function validPort(port: number): boolean {
+	return port >= 1 && port <= 65_535;
+}

package/evaluation-fixtures/tier1/boundary-error/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { validPort } from './defect';
+test('rejects the first invalid port', () => expect(validPort(65_536)).toBe(false));

package/evaluation-fixtures/tier1/boundary-error/environment/defect.ts

@@ -0,0 +1,3 @@
+export function validPort(port: number): boolean {
+	return port >= 1 && port <= 65_536;
+}

package/evaluation-fixtures/tier1/boundary-error/instruction.md

@@ -0,0 +1 @@
+Inspect the upper-bound validation in `environment/defect.ts` and identify the invalid value it permits.

package/evaluation-fixtures/tier1/boundary-error/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"boundary-error","source":"curated","split":"train","category":"boundary-error","protected":true,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"ac9cd9f603d86640900fe3af16e7373a322eb5deff009b72bdf4239df13661b8"}

package/evaluation-fixtures/tier1/branch-swap/environment/baseline.ts

@@ -0,0 +1,3 @@
+export function accessLabel(admin: boolean): string {
+	return admin ? 'admin' : 'user';
+}

package/evaluation-fixtures/tier1/branch-swap/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { accessLabel } from './defect';
+test('labels administrators', () => expect(accessLabel(true)).toBe('admin'));

package/evaluation-fixtures/tier1/branch-swap/environment/defect.ts

@@ -0,0 +1,3 @@
+export function accessLabel(admin: boolean): string {
+	return admin ? 'user' : 'admin';
+}

package/evaluation-fixtures/tier1/branch-swap/instruction.md

@@ -0,0 +1 @@
+Inspect `environment/defect.ts` for swapped conditional branches and report the incorrect results.

package/evaluation-fixtures/tier1/branch-swap/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"branch-swap","source":"curated","split":"train","category":"branch-swap","protected":false,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"f9b09fb7fa980fd94e1603a0e26e1ef43cd30a1d91211a195933913847943f23"}

package/evaluation-fixtures/tier1/curated-off-by-one/environment/baseline.ts

@@ -0,0 +1,3 @@
+export function pageStart(page: number, size: number): number {
+	return page * size;
+}

package/evaluation-fixtures/tier1/curated-off-by-one/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { pageStart } from './defect';
+test('first page begins at zero', () => expect(pageStart(0, 20)).toBe(0));

package/evaluation-fixtures/tier1/curated-off-by-one/environment/defect.ts

@@ -0,0 +1,3 @@
+export function pageStart(page: number, size: number): number {
+	return page * size + 1;
+}

package/evaluation-fixtures/tier1/curated-off-by-one/instruction.md

@@ -0,0 +1 @@
+Review the pagination boundary in `environment/defect.ts` and identify the page that is skipped.

package/evaluation-fixtures/tier1/curated-off-by-one/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"curated-off-by-one","source":"curated","split":"train","category":"curated-off-by-one","protected":true,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"2894129b530e0624dc1739ee7910839d261728a8655474b0103b8c98a7ea9f81"}

package/evaluation-fixtures/tier1/guard-removal/environment/baseline.ts

@@ -0,0 +1,3 @@
+export function normalized(value: string | undefined): string {
+	return value?.trim().toLowerCase() ?? '';
+}

package/evaluation-fixtures/tier1/guard-removal/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { normalized } from './defect';
+test('handles missing input', () => expect(normalized(undefined)).toBe(''));

package/evaluation-fixtures/tier1/guard-removal/environment/defect.ts

@@ -0,0 +1,3 @@
+export function normalized(value: string | undefined): string {
+	return value!.trim().toLowerCase();
+}

package/evaluation-fixtures/tier1/guard-removal/instruction.md

@@ -0,0 +1 @@
+Find the removed input guard in `environment/defect.ts` and state which input now fails.

package/evaluation-fixtures/tier1/guard-removal/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"guard-removal","source":"curated","split":"train","category":"guard-removal","protected":true,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"7943c549ab0c9ce6503ca8df45a9430ae0a431094b4e358be43da1cfb63a1f7d"}

package/evaluation-fixtures/tier1/injection-prone-string/environment/baseline.ts

@@ -0,0 +1,3 @@
+export function shellCommand(filename: string): string {
+	return `cat -- ${filename.replace(/[^A-Za-z0-9._-]/g, '')}`;
+}

package/evaluation-fixtures/tier1/injection-prone-string/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { shellCommand } from './defect';
+test('does not interpolate shell syntax', () => expect(shellCommand('a; rm -rf x')).not.toContain(';'));

package/evaluation-fixtures/tier1/injection-prone-string/environment/defect.ts

@@ -0,0 +1,3 @@
+export function shellCommand(filename: string): string {
+	return `cat ${filename}`;
+}

package/evaluation-fixtures/tier1/injection-prone-string/instruction.md

@@ -0,0 +1 @@
+Find the injection-prone command construction in `environment/defect.ts` and identify the unsafe input boundary.

package/evaluation-fixtures/tier1/injection-prone-string/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"injection-prone-string","source":"curated","split":"train","category":"injection-prone-string","protected":true,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"0f45604dbdd251aaec2af3f82afeec10b2242b16a88d86087994918796a657f4"}

package/evaluation-fixtures/tier1/missing-auth-check/environment/baseline.ts

@@ -0,0 +1,4 @@
+export function deleteAccount(actorRole: string, accountId: string): string {
+	if (actorRole !== 'admin') throw new Error('forbidden');
+	return `deleted:${accountId}`;
+}

package/evaluation-fixtures/tier1/missing-auth-check/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { deleteAccount } from './defect';
+test('blocks ordinary users', () => expect(() => deleteAccount('user', '42')).toThrow());

package/evaluation-fixtures/tier1/missing-auth-check/environment/defect.ts

@@ -0,0 +1,3 @@
+export function deleteAccount(_actorRole: string, accountId: string): string {
+	return `deleted:${accountId}`;
+}

package/evaluation-fixtures/tier1/missing-auth-check/instruction.md

@@ -0,0 +1 @@
+Review `environment/defect.ts` for a missing authorization check and state who can perform the protected action.

package/evaluation-fixtures/tier1/missing-auth-check/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"missing-auth-check","source":"curated","split":"train","category":"missing-auth-check","protected":true,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"9b6d8ff362df8aad25d3c579b10e0fa202e923a0565cf5d18d4c4d1da75af482"}

package/evaluation-fixtures/tier1/missing-await/environment/baseline.ts

@@ -0,0 +1,7 @@
+export async function saveThenRead(
+	save: () => Promise<void>,
+	read: () => string,
+): Promise<string> {
+	await save();
+	return read();
+}

package/evaluation-fixtures/tier1/missing-await/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { saveThenRead } from './defect';
+test('waits for save', async () => { let value = ''; expect(await saveThenRead(async () => { await Promise.resolve(); value = 'saved'; }, () => value)).toBe('saved'); });

package/evaluation-fixtures/tier1/missing-await/environment/defect.ts

@@ -0,0 +1,4 @@
+export async function saveThenRead(save: () => Promise<void>, read: () => string): Promise<string> {
+	save();
+	return read();
+}

package/evaluation-fixtures/tier1/missing-await/instruction.md

@@ -0,0 +1 @@
+Find the missing await in `environment/defect.ts` and explain the ordering failure.

package/evaluation-fixtures/tier1/missing-await/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"missing-await","source":"curated","split":"train","category":"missing-await","protected":true,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"d444592ce7d955882de61601bfe414be54c616f6c441ea735450e6716bc837fd"}

package/evaluation-fixtures/tier1/mutation-off-by-one/environment/baseline.ts

@@ -0,0 +1,3 @@
+export function last<T>(items: T[]): T | undefined {
+	return items[items.length - 1];
+}

package/evaluation-fixtures/tier1/mutation-off-by-one/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { last } from './defect';
+test('returns the final item', () => expect(last([1, 2])).toBe(2));

package/evaluation-fixtures/tier1/mutation-off-by-one/environment/defect.ts

@@ -0,0 +1,3 @@
+export function last<T>(items: T[]): T | undefined {
+	return items[items.length];
+}

package/evaluation-fixtures/tier1/mutation-off-by-one/instruction.md

@@ -0,0 +1 @@
+Find the boundary defect in `environment/defect.ts`. Report whether the implementation is safe and explain the exact failing case.

package/evaluation-fixtures/tier1/mutation-off-by-one/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"mutation-off-by-one","source":"curated","split":"train","category":"mutation-off-by-one","protected":true,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"9cec6a89ff87b1b85691313a2430916e3ebf393d532644b3f82faa10fe0203cd"}

package/evaluation-fixtures/tier1/null-substitution/environment/baseline.ts

@@ -0,0 +1,3 @@
+export function displayName(name: string | null): string {
+	return name === null ? '' : name;
+}

package/evaluation-fixtures/tier1/null-substitution/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { displayName } from './defect';
+test('preserves a real name', () => expect(displayName('Ada')).toBe('Ada'));

package/evaluation-fixtures/tier1/null-substitution/environment/defect.ts

@@ -0,0 +1,3 @@
+export function displayName(name: string | null): string {
+	return name === null ? '' : null as unknown as string;
+}

package/evaluation-fixtures/tier1/null-substitution/instruction.md

@@ -0,0 +1 @@
+Inspect `environment/defect.ts` for an incorrect null substitution and report the observable regression.

package/evaluation-fixtures/tier1/null-substitution/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"null-substitution","source":"curated","split":"train","category":"null-substitution","protected":true,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"5a335a9750981580e3feb6449d348f5111ff84a77058b8b0a50b57caf8ba9a6c"}

package/evaluation-fixtures/tier1/operator-swap/environment/baseline.ts

@@ -0,0 +1,3 @@
+export function canRetry(attempt: number, maximum: number): boolean {
+	return attempt < maximum;
+}

package/evaluation-fixtures/tier1/operator-swap/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { canRetry } from './defect';
+test('allows attempts below the maximum', () => expect(canRetry(1, 3)).toBe(true));

package/evaluation-fixtures/tier1/operator-swap/environment/defect.ts

@@ -0,0 +1,3 @@
+export function canRetry(attempt: number, maximum: number): boolean {
+	return attempt > maximum;
+}

package/evaluation-fixtures/tier1/operator-swap/instruction.md

@@ -0,0 +1 @@
+Review the comparison in `environment/defect.ts` and identify the incorrect operator.

package/evaluation-fixtures/tier1/operator-swap/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"operator-swap","source":"curated","split":"train","category":"operator-swap","protected":false,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"613431a4462af61ac236985724be5879b8b1d713ea090f74a8713011378b781e"}

package/evaluation-fixtures/tier1/side-effect-deletion/environment/baseline.ts

@@ -0,0 +1,4 @@
+export function record(items: string[], value: string): number {
+	items.push(value);
+	return items.length;
+}

package/evaluation-fixtures/tier1/side-effect-deletion/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { record } from './defect';
+test('records the value', () => { const items: string[] = []; record(items, 'x'); expect(items).toEqual(['x']); });

package/evaluation-fixtures/tier1/side-effect-deletion/environment/defect.ts

@@ -0,0 +1,3 @@
+export function record(items: string[], value: string): number {
+	return items.length + 1;
+}

package/evaluation-fixtures/tier1/side-effect-deletion/instruction.md

@@ -0,0 +1 @@
+Identify the missing side effect in `environment/defect.ts` and explain how callers observe it.

package/evaluation-fixtures/tier1/side-effect-deletion/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"side-effect-deletion","source":"curated","split":"train","category":"side-effect-deletion","protected":false,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"d86bd1df176087b1157480043098782ea671553a713015ec734cd573fa23be88"}

package/evaluation-fixtures/tier1/swallowed-error/environment/baseline.ts

@@ -0,0 +1,3 @@
+export async function required(work: () => Promise<void>): Promise<void> {
+	await work();
+}

package/evaluation-fixtures/tier1/swallowed-error/environment/defect.test.ts

@@ -0,0 +1,3 @@
+import { expect, test } from 'bun:test';
+import { required } from './defect';
+test('propagates failures', async () => expect(required(async () => { throw new Error('boom'); })).rejects.toThrow('boom'));

package/evaluation-fixtures/tier1/swallowed-error/environment/defect.ts

@@ -0,0 +1,3 @@
+export async function required(work: () => Promise<void>): Promise<void> {
+	try { await work(); } catch { return; }
+}

package/evaluation-fixtures/tier1/swallowed-error/instruction.md

@@ -0,0 +1 @@
+Inspect the error handling in `environment/defect.ts` and report why failures become invisible.

package/evaluation-fixtures/tier1/swallowed-error/manifest.json

@@ -0,0 +1 @@
+{"v":1,"id":"swallowed-error","source":"curated","split":"train","category":"swallowed-error","protected":true,"instructionPath":"instruction.md","environment":{"kind":"fixture","path":"environment"},"scorer":{"kind":"builtin","argv":["tier1-defect"],"timeoutMs":30000,"scoreRange":[0,1]},"provenance":{"origin":"opencode-swarm#1820","license":"MIT"},"contentHash":"4889fcc71918cfc11aec97258b1be29158e3e353a029446aad848dce4cf5583f"}

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "opencode-swarm",
-	"version": "7.115.0",
+	"version": "7.116.0",
 	"description": "Architect-centric agentic swarm plugin for OpenCode - hub-and-spoke orchestration with SME consultation, code generation, and QA review",
 	"main": "dist/index.js",
 	"types": "dist/index.d.ts",
@@ -41,6 +41,7 @@
 		"dist",
 		"dist/lang/grammars",
 		"binaries",
+		"evaluation-fixtures",
 		".opencode/skills/brainstorm",
 		".opencode/skills/specify",
 		".opencode/skills/clarify-spec",