@duckmind/dm-darwin-x64 0.48.5 → 0.49.6

package/extensions/dm-subagents/.m/artifacts/shell/shell_output_1783846154938_6a41cefee0bf.txt

@@ -0,0 +1,1336 @@
+
+> dm-subagents@0.33.1 test:unit
+> node --experimental-strip-types --test test/unit/*.test.ts
+
+▶ acceptance gates
+  ✔ infers different policies for reviewer, writer, async writer, and dynamic contexts (0.833833ms)
+  ✔ explicit acceptance can strengthen inferred policy (0.07525ms)
+  ✔ formats a standardized child prompt section (0.163666ms)
+  ✔ parses acceptance-report fences and ignores unrelated json fences (1.436625ms)
+  ✔ parses acceptance reports from json-family fences (0.142167ms)
+  ✔ strips trailing json-family reports after earlier unrelated json fences (0.074ms)
+  ✔ unwraps acceptance-report wrapper objects (0.078584ms)
+  ✔ reports field-level validation errors for malformed acceptance-report fields (0.107958ms)
+  ✔ explicit none disables inferred gates when a reason is present (0.060041ms)
+  ✔ checked mode rejects missing required evidence (78.462417ms)
+  ✔ surfaces parse validation details in acceptance failure messages (3.578167ms)
+  ✔ checked mode rejects not-satisfied required criteria (20.845792ms)
+  ✔ verified mode records runtime command success and failure separately from child command claims (96.770791ms)
+  ✔ reviewed mode records no-blocker and blocker reviewer outcomes (88.787ms)
+  ✔ does not make explicit checked acceptance an explicit reviewed blocker when inference recommends review (15.421875ms)
+  ✔ does not mark reviewed without an independent reviewer result (43.035208ms)
+  ✔ zero-child aggregate reports do not fabricate required evidence (28.037083ms)
+  ✔ validates invalid disable and verify shapes (0.314459ms)
+✔ acceptance gates (379.394958ms)
+▶ builtin agent disabling
+  ✔ filters a per-agent disabled builtin from runtime discovery while keeping it in discoverAgentsAll (202.310292ms)
+  ✔ surfaces malformed disabled overrides instead of silently ignoring them (5.823125ms)
+  ✔ bulk disableBuiltins hides builtins at runtime and marks them disabled in discoverAgentsAll (11.062334ms)
+  ✔ an explicit user override opts a builtin out of user-scope bulk disable (6.122125ms)
+  ✔ project disableBuiltins false re-enables builtins hidden by user bulk disable (3.0505ms)
+  ✔ project bulk disable beats user per-agent re-enable overrides (4.771333ms)
+  ✔ surfaces malformed disableBuiltins values instead of silently ignoring them (7.411ms)
+  ✔ hides disabled builtins from agent-facing management list output (241.1095ms)
+  ✔ buildBuiltinOverrideConfig emits disabled false when re-enabling a builtin (0.783334ms)
+✔ builtin agent disabling (483.041166ms)
+▶ agent eject/disable/enable/reset management actions
+  ▶ eject
+    ✔ copies a bundled builtin to user scope verbatim and shadows the builtin (186.329375ms)
+    ✔ ejects to project scope when agentScope is project (8.040167ms)
+    ✔ copies a package agent that shadows a builtin by runtime precedence (14.600292ms)
+    ✔ refuses invalid management scopes without writing user files (0.316791ms)
+    ✔ refuses to eject when a custom agent already exists (6.214292ms)
+    ✔ refuses to eject an unknown agent (9.25275ms)
+  ✔ eject (227.48925ms)
+  ▶ disable
+    ✔ hides a builtin from runtime discovery via a user settings override (35.629167ms)
+    ✔ merges disabled into an existing override without dropping other fields (18.348209ms)
+    ✔ hides a disabled agent from agent-facing list output (226.548958ms)
+    ✔ writes a project-scoped override when agentScope is project (8.423791ms)
+    ✔ refuses to disable an unknown agent (6.950917ms)
+    ✔ also disables a custom agent via a settings override (10.795ms)
+    ✔ also disables a package agent via a settings override (20.835791ms)
+  ✔ disable (327.794875ms)
+  ▶ enable
+    ✔ restores a previously disabled builtin to runtime discovery (21.332375ms)
+    ✔ preserves other override fields when removing the disabled flag (12.120459ms)
+    ✔ reports already enabled and makes no changes when nothing is disabled (6.468917ms)
+    ✔ points to the disabling scope when enabling the wrong scope (8.465625ms)
+    ✔ refuses to enable an unknown agent (5.288959ms)
+  ✔ enable (53.8105ms)
+  ▶ reset
+    ✔ deletes a custom shadow file and restores the bundled builtin (15.558583ms)
+    ✔ removes a settings override and restores the pristine builtin (14.409041ms)
+    ✔ removes both a custom file and a settings override in one reset (25.189042ms)
+    ✔ reports a no-op when there is nothing to reset in the target scope (3.16725ms)
+    ✔ points to delete for a custom agent with no bundled default (3.432708ms)
+    ✔ refuses to reset an unknown agent (5.846375ms)
+  ✔ reset (67.735125ms)
+✔ agent eject/disable/enable/reset management actions (677.107791ms)
+▶ agent permission frontmatter
+  ✔ preserves nested permission YAML blocks through discovery and serialization (13.633333ms)
+✔ agent permission frontmatter (13.969666ms)
+▶ agent frontmatter defaultContext
+  ✔ serializes defaultContext into agent frontmatter (0.115667ms)
+  ✔ parses defaultContext from discovered agent frontmatter (5.840792ms)
+  ✔ loads packaged planner, worker, and oracle with fork defaultContext (118.820416ms)
+✔ agent frontmatter defaultContext (124.88975ms)
+▶ chain discovery
+  ✔ prefers same-scope .chain.json over .chain.md for the same runtime name (4.368417ms)
+✔ chain discovery (4.415375ms)
+▶ package-provided agents and chains
+  ✔ discovers package agents and chains from installed package manifests (7.148917ms)
+  ✔ loads packages referenced from DM settings (3.5345ms)
+  ✔ discovers project package agents when cwd is nested below the project root (13.311291ms)
+  ✔ does not register legacy skill files from broad package agent roots (8.931ms)
+  ✔ keeps package definitions below user and project overrides (19.177458ms)
+  ✔ does not allow management updates to package agents (5.755417ms)
+✔ package-provided agents and chains (58.091167ms)
+▶ agent frontmatter completionGuard
+  ✔ serializes disabled completion guard into agent frontmatter (0.094583ms)
+  ✔ omits enabled completion guard from serialized frontmatter (0.059917ms)
+  ✔ parses completionGuard from discovered agent frontmatter (1.849167ms)
+✔ agent frontmatter completionGuard (2.05275ms)
+▶ agent frontmatter maxSubagentDepth
+  ✔ serializes maxSubagentDepth into agent frontmatter (0.057833ms)
+  ✔ parses maxSubagentDepth from discovered agent frontmatter (1.685708ms)
+✔ agent frontmatter maxSubagentDepth (1.775584ms)
+▶ agent frontmatter fallbackModels
+  ✔ serializes fallbackModels into agent frontmatter (0.0535ms)
+  ✔ parses fallbackModels from discovered agent frontmatter (4.577167ms)
+✔ agent frontmatter fallbackModels (4.728458ms)
+▶ agent frontmatter systemPromptMode
+  ✔ serializes systemPromptMode into agent frontmatter (0.309ms)
+  ✔ parses systemPromptMode from discovered agent frontmatter (4.15375ms)
+✔ agent frontmatter systemPromptMode (4.579167ms)
+▶ agent frontmatter prompt inheritance flags
+  ✔ serializes inheritProjectContext and inheritSkills into agent frontmatter (0.163041ms)
+  ✔ parses inheritProjectContext and inheritSkills from discovered agent frontmatter (10.764041ms)
+✔ agent frontmatter prompt inheritance flags (11.032625ms)
+▶ agent frontmatter subagentOnlyExtensions
+  ✔ serializes subagentOnlyExtensions into agent frontmatter (0.238875ms)
+  ✔ parses subagentOnlyExtensions from discovered agent frontmatter (2.970417ms)
+✔ agent frontmatter subagentOnlyExtensions (3.293ms)
+▶ agent frontmatter prompt assembly defaults
+  ✔ defaults ordinary agents to replace mode with no inherited context or skills (1.789292ms)
+  ✔ builtin agents inherit project context by default (3.797791ms)
+  ✔ bundled agents all have explicit tool allowlists (3.948458ms)
+  ✔ worker and delegate include the child-facing supervisor tool (4.182792ms)
+  ✔ defaults delegate to append mode with inherited project context (1.571792ms)
+✔ agent frontmatter prompt assembly defaults (15.365375ms)
+▶ packaged agent and chain discovery
+  ✔ recursively discovers nested project agents while keeping chain files separate (6.958959ms)
+  ✔ registers packaged agents by runtime name and serializes local name plus package (3.176375ms)
+  ✔ recursively discovers packaged chains by runtime name and preserves package on serialize (6.17575ms)
+  ✔ keeps packaged and un-packaged runtime names distinct while preserving un-packaged precedence (2.035333ms)
+  ✔ parses packaged chains directly from serializer helpers (0.095125ms)
+  ✔ normalizes package frontmatter consistently for agents and chains (4.674542ms)
+  ✔ skips invalid package frontmatter that cannot be normalized (2.818083ms)
+✔ packaged agent and chain discovery (26.086333ms)
+▶ project agent directory discovery
+  ✔ discovers project agents from both .agents and .dm/agents (2.255667ms)
+  ✔ does not register legacy project skill files as agents (2.25675ms)
+  ✔ does not register user SKILL.md files as agents (4.446417ms)
+  ✔ prefers .dm/agents over .agents on project agent name collisions (2.152958ms)
+  ✔ uses the project root for the canonical project agent dir even when only .agents exists (8.777875ms)
+  ✔ discovers project chains from .dm/chains (3.30725ms)
+  ✔ prefers project .dm/chains over user chains on name collisions (3.2065ms)
+✔ project agent directory discovery (26.533875ms)
+▶ agent management config parsing
+  ✔ surfaces JSON parse errors for create config strings (1.850916ms)
+  ✔ surfaces JSON parse errors for update config strings (1.02775ms)
+  ✔ creates, gets, updates, and deletes a packaged agent by runtime name (142.836ms)
+  ✔ rejects package values that cannot be normalized (2.100667ms)
+  ✔ creates and updates packaged chains while preserving packaged step names (209.142666ms)
+  ✔ creates and updates agents with tool budgets (15.234208ms)
+  ✔ rejects invalid tool budget management config (18.178958ms)
+  ✔ creates agents with completion guard disabled (6.786709ms)
+  ✔ rejects non-boolean completion guard config (8.6825ms)
+  ✔ creates agents with subagent-only extensions (13.717834ms)
+  ✔ does not serialize settings overrides into custom agent frontmatter during updates (5.200041ms)
+  ✔ preserves explicit default-like frontmatter that blocks settings overrides during updates (4.702875ms)
+  ✔ updates JSON chain descriptions without rewriting them as markdown (23.256709ms)
+  ✔ renames and repackages JSON chains while preserving JSON format and extension (5.641792ms)
+  ✔ gets dynamic JSON chain details and lists invalid chain diagnostics (12.408542ms)
+  ✔ reports builtin runtime-loaded model mappings from current session state (4.5295ms)
+  ✔ reports override source and disabled builtin state in runtime model mappings (2.505833ms)
+  ✔ rejects unknown builtin filters for runtime model mappings (0.209209ms)
+  ✔ creates delegate with its builtin prompt defaults (6.25725ms)
+  ✔ lists proactive skill subagent suggestions from repeated configured skill use (4.094125ms)
+  ✔ can disable proactive skill subagent suggestions in config (2.804042ms)
+✔ agent management config parsing (492.213917ms)
+▶ parseMemoryFrontmatter
+  ✔ parses a project scope block (0.799791ms)
+  ✔ parses a user scope with nested path and strips quotes (0.26725ms)
+  ✔ parses an inline object memory block (0.301416ms)
+  ✔ rejects unknown scopes (0.902333ms)
+  ✔ rejects a missing path (0.279583ms)
+  ✔ rejects a missing scope (0.392958ms)
+  ✔ treats empty or absent input as no memory config (1.848208ms)
+✔ parseMemoryFrontmatter (5.763875ms)
+▶ agentHasWriteTools
+  ✔ inherits write capability when tools are unset (0.659042ms)
+  ✔ detects edit, write, and bash as write tools (0.123125ms)
+  ✔ treats read-only and mcp-only tool sets as non-write (0.1225ms)
+✔ agentHasWriteTools (1.083917ms)
+▶ resolveMemoryDir
+  ✔ resolves a simple and nested path under the root (2.685292ms)
+  ✔ rejects empty paths (2.247917ms)
+  ✔ rejects dot and parent segments (0.685625ms)
+  ✔ rejects absolute and Windows drive-like paths (3.113292ms)
+  ✔ rejects a symlinked ancestor before prompting a first write (9.171084ms)
+  ✔ rejects a memory dir that is a symlink escaping the root (1.147208ms)
+✔ resolveMemoryDir (19.878833ms)
+▶ readMemoryFile
+  ✔ returns null when no memory file exists (15.480084ms)
+  ✔ reads contents and reports byte capping (2.914958ms)
+  ✔ flags byte-capped contents (1.384083ms)
+  ✔ does not return more than the memory byte cap (2.035ms)
+  ✔ rejects a symlinked memory file that escapes the memory dir (0.604875ms)
+✔ readMemoryFile (22.920542ms)
+▶ buildAgentMemoryInjection
+  ✔ returns empty when the agent has no memory scope (0.440541ms)
+  ✔ injects a read-write block with contents for a project scope (1.5095ms)
+  ✔ injects a creation prompt when a read-write agent has no memory file yet (0.731542ms)
+  ✔ injects a read-only block for agents without write tools (4.109791ms)
+  ✔ injects nothing for a read-only agent with no memory file yet (0.797833ms)
+  ✔ resolves user scope under the agent dir, separate from the owner memory system (2.953292ms)
+  ✔ returns empty for project scope when no project root is found (3.051959ms)
+  ✔ returns empty when the memory path is unsafe (2.078334ms)
+  ✔ returns empty when the memory file is an escaping symlink (2.565791ms)
+  ✔ caps memory contents to the first N lines without claiming a byte cap (1.696791ms)
+✔ buildAgentMemoryInjection (20.196417ms)
+▶ agent memory frontmatter round-trip
+  ✔ parses memory frontmatter during discovery and keeps it out of extraFields (8.429708ms)
+  ✔ serializes memory back into frontmatter (0.202292ms)
+  ✔ ignores a malformed memory block without dropping the agent (13.518833ms)
+✔ agent memory frontmatter round-trip (22.236041ms)
+▶ agent memory in management detail
+  ✔ surfaces the memory scope in the get action (132.843875ms)
+✔ agent memory in management detail (132.883417ms)
+▶ builtin agent overrides
+  ✔ bundled builtin agents inherit the default model (158.704041ms)
+  ✔ applies subagents.defaultModel to builtin agents with explicit overrides winning (4.838167ms)
+  ✔ prefers project subagents.defaultModel over user defaultModel (2.527792ms)
+  ✔ applies subagents.defaultModel to custom agents without a frontmatter model (2.8905ms)
+  ✔ applies user settings overrides to builtin agents (3.448875ms)
+  ✔ globally disables builtin thinking suffix defaults from user settings (17.148042ms)
+  ✔ lets an explicit same-scope thinking override opt back in when global thinking is disabled (3.63275ms)
+  ✔ lets project settings disable builtin thinking even when user overrides request it (8.000125ms)
+  ✔ surfaces malformed subagent default model settings (2.301875ms)
+  ✔ surfaces malformed global thinking settings (1.975084ms)
+  ✔ prefers project settings overrides over user settings overrides (11.192ms)
+  ✔ does not apply project settings overrides when scope is user (3.880375ms)
+  ✔ does not apply user settings overrides when scope is project (2.213333ms)
+  ✔ does not read malformed out-of-scope settings files (2.334625ms)
+  ✔ frontmatter wins per-field over agentOverrides for a shadowing project agent (7.813416ms)
+  ✔ fills in unset fields on a custom project agent from project agentOverrides (3.080625ms)
+  ✔ fills in unset fields on a custom user agent from user agentOverrides (6.23525ms)
+  ✔ applies user agentOverrides to a custom project agent when project settings have no entry (2.858792ms)
+  ✔ prefers project agentOverrides over user agentOverrides on a custom project agent (2.525958ms)
+  ✔ keeps explicit custom frontmatter fields over matching agentOverrides (5.825291ms)
+  ✔ leaves a custom agent untouched when no agentOverrides entry matches its name (15.412792ms)
+  ✔ disableBuiltins does not disable custom agents (3.215292ms)
+  ✔ does not create a settings file when removing a non-existent override (4.348917ms)
+  ✔ surfaces malformed settings files instead of silently ignoring them (2.916583ms)
+  ✔ surfaces settings read failures without mislabeling them as parse errors (1.915ms)
+  ✔ surfaces malformed builtin override entries instead of silently ignoring them (2.069583ms)
+  ✔ surfaces malformed completion guard override values (1.851833ms)
+  ✔ builds false sentinels when an override clears builtin fields (0.352208ms)
+✔ builtin agent overrides (286.461375ms)
+▶ resolveExecutionAgentScope
+  ✔ defaults to both when scope is omitted (0.283875ms)
+  ✔ passes through explicit scopes (0.050583ms)
+  ✔ falls back to both for invalid scopes (0.04825ms)
+✔ resolveExecutionAgentScope (0.81475ms)
+▶ mergeAgentsForScope
+  ✔ returns project agents when scope is project (2.596542ms)
+  ✔ returns user agents when scope is user (0.066542ms)
+  ✔ prefers project agents on name collisions when scope is both (0.047459ms)
+  ✔ keeps agents from both scopes when names are distinct (0.076125ms)
+  ✔ includes builtin agents when no user or project override exists (0.045167ms)
+  ✔ user agents override builtins with the same name (0.042667ms)
+  ✔ package agents override builtins but not user or project agents (0.057167ms)
+  ✔ project agents override builtins with the same name (0.041125ms)
+✔ mergeAgentsForScope (5.364ms)
+▶ project-local artifact paths
+  ✔ places generated subagent files under .dm-subagents for a project cwd (0.298042ms)
+  ✔ keeps the session artifact fallback when no project cwd is available (0.073125ms)
+✔ project-local artifact paths (0.83975ms)
+▶ async runner execution
+  ✔ places detached runner stdio logs in the async run directory (0.506833ms)
+  ✔ omits runner log paths when asyncDir is unavailable (0.051916ms)
+  ✔ resolves async step tool budgets with step over run over agent over config precedence (1.571334ms)
+  ✔ uses agent tool budget before config default when no run override exists (0.127625ms)
+  ✔ uses config default when no step, run, or agent budget exists (0.102667ms)
+✔ async runner execution (2.764584ms)
+▶ async interrupt action
+  ✔ queues steering for a running async child (2.507625ms)
+  ✔ queues steering for a running async child by directory (2.061375ms)
+  ✔ queues steering for a pending indexed async child (1.822625ms)
+  ✔ interrupts a running async run resolved from disk after in-memory tracking is gone (1.268958ms)
+  ✔ reports success and writes the portable request when the signal is unavailable (0.838958ms)
+  ✔ does not report success for stale running status with a dead pid (1.1585ms)
+✔ async interrupt action (10.19575ms)
+▶ async permission forwarding session identity
+  ✔ uses the parent session id for permission forwarding instead of the async status identity (1.460375ms)
+  ✔ consumes bounded dynamic fanout indexes before later static forked steps (3.777834ms)
+  ✔ applies thinking overrides to async fallback candidates (0.177583ms)
+✔ async permission forwarding session identity (5.781667ms)
+▶ async resume lookup
+  ✔ resolves a completed single-child run from persisted status (1.635042ms)
+  ✔ rejects ambiguous run id prefixes (0.976625ms)
+  ✔ rejects path-like ids and directories outside the async root (0.21825ms)
+  ✔ rejects non-jsonl session files before reviving (0.595542ms)
+  ✔ rejects malformed result metadata before using session fields (0.524833ms)
+  ✔ rejects malformed status session ids (0.575417ms)
+  ✔ returns a live intercom target for a running child (0.533917ms)
+  ✔ revives a completed child by index while a sibling async child is still running (0.698375ms)
+  ✔ rejects pending indexed children in still-running async runs (0.658334ms)
+  ✔ resolves a completed multi-child run when an index and per-child session file are available (0.763625ms)
+  ✔ frames the revived follow-up with original run context (0.11625ms)
+✔ async resume lookup (7.789042ms)
+▶ writeAtomicJson
+  ✔ retries transient rename failures before replacing the target (0.623833ms)
+  ✔ uses longer default retries for transient Windows rename locks (0.098167ms)
+  ✔ throws non-retryable rename failures without retrying (0.178667ms)
+  ✔ cleans up the temp file after retryable failures are exhausted (0.306792ms)
+✔ writeAtomicJson (1.656333ms)
+▶ chain append requests
+  ✔ persists append requests for running async chains and records pending status (5.260667ms)
+  ✔ reads pending append requests without consuming them and reports reserved output names (1.364167ms)
+  ✔ rejects terminal and non-chain async runs (5.837083ms)
+  ✔ extends async chain status, parallel groups, and workflow graph (0.763167ms)
+✔ chain append requests (14.183084ms)
+▶ async chain root attachment
+  ✔ imports an already completed async child result (1.64775ms)
+  ✔ waits for a running async child to write its terminal result (28.525291ms)
+  ✔ imports a failed root as a failed first chain step (2.61125ms)
+  ✔ fails a terminal root that never produced a result file (1.919708ms)
+  ✔ stops waiting when the parent timeout aborts an attached root (12.649375ms)
+✔ async chain root attachment (48.210167ms)
+▶ chain serializer
+  ✔ round-trips step outputMode (0.689625ms)
+  ✔ round-trips phase, label, as, and path-based outputSchema (0.082125ms)
+  ✔ round-trips markdown chain toolBudget (0.356041ms)
+  ✔ rejects invalid markdown chain toolBudget (6.087ms)
+  ✔ rejects inline outputSchema values in markdown chains (0.219334ms)
+  ✔ rejects invalid dynamic JSON chains with useful diagnostics (10.373625ms)
+  ✔ serializes JSON chains back to JSON (7.7245ms)
+  ✔ parses declarative JSON chains with dynamic fanout toolBudget (1.855916ms)
+  ✔ rejects invalid JSON chain toolBudget (3.493125ms)
+  ✔ parses declarative JSON chains with dynamic fanout (0.847875ms)
+  ✔ parses and validates acceptance in JSON chains (0.50625ms)
+✔ chain serializer (37.89775ms)
+▶ createChildTranscriptWriter
+  ✔ writes a versioned initial user message record with run metadata (1.448ms)
+  ✔ omits childIndex when not provided (0.884417ms)
+  ✔ writes message records for message_end and tool_result_end events (4.835208ms)
+  ✔ writes tool_start and tool_end records and ignores unhandled event types (0.519291ms)
+  ✔ skips blank stdout/stderr lines and splits multi-line stderr text (0.432625ms)
+  ✔ truncates after the byte cap and stops writing further records without erroring (0.385792ms)
+  ✔ reserves space for a truncation marker before accepting another record (0.321916ms)
+  ✔ reports an init error and drops subsequent writes when the transcript path cannot be created (0.233458ms)
+✔ createChildTranscriptWriter (9.605709ms)
+▶ attachPostExitStdioGuard
+  ✔ reports whether a termination signal was actually delivered (0.30975ms)
+  ✔ does not delay a clean exit (15.452958ms)
+  ✔ cuts off a silent grandchild with the idle timer (1537.668958ms)
+  ✔ cuts off a chatty grandchild with the hard timer (2026.334834ms)
+✔ attachPostExitStdioGuard (3580.335292ms)
+▶ resolveCompletionBatchConfig
+  ✔ applies defaults when no config is provided (1.079708ms)
+  ✔ override wins over global config (0.249667ms)
+  ✔ rejects invalid booleans and non-positive or fractional integers (0.132208ms)
+  ✔ ignores invalid overrides instead of masking valid global config (0.094ms)
+✔ resolveCompletionBatchConfig (2.530375ms)
+▶ createCompletionBatcher
+  ✔ emits immediately when batching is disabled (0.492708ms)
+  ✔ coalesces a burst into one grouped emit when the debounce fires (0.359917ms)
+  ✔ emits at the max-wait cap even when items keep arriving (1.003708ms)
+  ✔ flush emits held items immediately and clears timers (0.200875ms)
+  ✔ dispose clears timers without emitting (0.204792ms)
+  ✔ creates a shorter straggler group for late siblings (0.24075ms)
+  ✔ starts a fresh normal group after the straggler window passes (0.159875ms)
+✔ createCompletionBatcher (2.977084ms)
+▶ buildCompletionKey
+  ✔ uses id as canonical key when present (0.620375ms)
+  ✔ builds deterministic fallback key when id is missing (0.217459ms)
+✔ buildCompletionKey (1.714208ms)
+▶ markSeenWithTtl
+  ✔ returns true only for duplicates within ttl (0.087292ms)
+✔ markSeenWithTtl (0.141417ms)
+▶ getGlobalSeenMap
+  ✔ returns the same map for the same global store key (0.073291ms)
+✔ getGlobalSeenMap (0.113917ms)
+✔ implementation task with no mutation triggers the completion guard (0.955417ms)
+✔ declared read-only builtin tools suppress implementation-word false positives (0.091584ms)
+✔ omitted, empty, bash, unknown, write, and MCP tool capabilities stay conservative (0.804416ms)
+✔ worker with mutating-capable tools still triggers when no mutation is observed (0.247125ms)
+✔ review-only, research, and framework output instructions do not expect mutation (2.388333ms)
+✔ worker implementation verbs win over investigative wording (0.182209ms)
+✔ worker edit intent covers common docs, config, and source tasks (0.433333ms)
+✔ edit and write tool calls count as mutation attempts (0.132125ms)
+✔ obvious mutating bash commands count as mutation attempts (1.175292ms)
+✔ implementation task with mutation attempts does not trigger (0.140791ms)
+▶ config directory resolution
+  ✔ falls back without importing the DM peer package at runtime (1.529166ms)
+  ✔ honors an explicitly provided DM module shape (0.22925ms)
+  ✔ honors DM package metadata without importing the peer package (0.851625ms)
+  ✔ uses an explicit DM package root before the process entrypoint walk (0.498208ms)
+  ✔ does not runtime-import the coding agent peer from shared utils (0.216042ms)
+✔ config directory resolution (3.996709ms)
+▶ control channel: request file
+  ✔ writes a parseable interrupt request, creating the inbox dir (2.206ms)
+  ✔ keeps the request type authoritative even for untyped callers (0.597542ms)
+  ✔ consumes a pending request exactly once and removes the file (0.742542ms)
+  ✔ removes a malformed request directory instead of firing forever (0.527125ms)
+  ✔ writes and consumes ordered steer requests (2.992584ms)
+  ✔ keeps steer request ids out of filesystem paths (1.731125ms)
+  ✔ does not deliver a steer request if another consumer removed it first (1.85475ms)
+  ✔ enqueues a steer request for a specific child inbox (2.22625ms)
+  ✔ rejects empty steer messages and invalid target indexes (0.84525ms)
+✔ control channel: request file (15.965458ms)
+▶ control channel: deliverInterruptRequest
+  ✔ writes the portable request and signals best-effort when kill succeeds (19.865958ms)
+  ✔ still writes the request when the OS signal throws ENOSYS (Windows) (0.942375ms)
+  ✔ surfaces non-portability signal failures and removes the stale request (1.4695ms)
+  ✔ skips signalling when no live pid is provided (1.498708ms)
+✔ control channel: deliverInterruptRequest (23.927ms)
+▶ control channel: watchAsyncControlInbox
+  ✔ fires on a request that arrived before the watcher started (1.765333ms)
+  ✔ fires once per request via the watch event and stops after dispose (2.175834ms)
+  ✔ delivers steer requests without firing interrupt (1.315416ms)
+✔ control channel: watchAsyncControlInbox (5.371667ms)
+▶ subagent control notice delivery
+  ✔ delivers async needs-attention notices immediately (5.639875ms)
+  ✔ queues foreground needs-attention notices until the same step is still actionable (27.938167ms)
+  ✔ drops queued foreground notices when the run finishes before delivery (37.596042ms)
+  ✔ drops queued foreground notices after the chain advances to another step (27.876041ms)
+✔ subagent control notice delivery (100.545917ms)
+▶ resolveRuntimeExtensionPath
+  ✔ resolves helpers from the package root for standalone and bundled modules (0.39ms)
+✔ resolveRuntimeExtensionPath (0.716041ms)
+▶ buildPiArgs session wiring
+  ✔ uses --session when sessionFile is provided (0.705209ms)
+  ✔ keeps fresh mode behavior (sessionDir + no session file) (0.097459ms)
+  ✔ emits explicit parent session env for permission forwarding (0.09575ms)
+  ✔ falls back to inherited parent session env for permission forwarding (0.230417ms)
+✔ buildPiArgs session wiring (1.280791ms)
+▶ buildPiArgs model wiring
+  ✔ uses --model for provider-qualified model ids (0.36775ms)
+  ✔ uses --model for bare model ids too (0.203375ms)
+  ✔ preserves thinking suffixes on model args (0.172459ms)
+  ✔ passes explicit thinking off through to the model arg (0.157916ms)
+  ✔ leaves provider-specific model suffixes untouched when thinking is disabled (0.2205ms)
+✔ buildPiArgs model wiring (1.395125ms)
+▶ buildPiArgs system prompt mode wiring
+  ✔ uses --append-system-prompt by default (0.943833ms)
+  ✔ uses --system-prompt when systemPromptMode=replace (0.628ms)
+  ✔ injects the subagent prompt runtime extension and env flags (0.19975ms)
+  ✔ passes tool budget through env (0.733125ms)
+  ✔ passes child intercom and orchestrator metadata through env (0.464209ms)
+  ✔ does not create a supervisor channel without an exact parent session id (0.207291ms)
+  ✔ emits explicit builtin tool allowlists (0.190583ms)
+  ✔ adds read to explicit tool allowlists when skills must be loaded lazily (0.149ms)
+  ✔ does not duplicate read in explicit tool allowlists for lazy skills (0.137875ms)
+  ✔ augments explicit builtin allowlists with selected direct MCP tool names (4.379875ms)
+  ✔ preserves no --tools for MCP-only agents (9.944542ms)
+  ✔ supports direct MCP server/tool filters (1.362166ms)
+  ✔ matches adapter prefix modes for direct MCP names (3.086959ms)
+  ✔ includes resource tools and respects excludeTools (1.138708ms)
+  ✔ falls back to explicit builtins when direct MCP cache or config is missing or invalid (3.550375ms)
+  ✔ resolves project MCP config from the child cwd and expands DM_CODING_AGENT_DIR (2.50175ms)
+  ✔ keeps tool extension paths when explicit extensions are allowlisted (2.015042ms)
+  ✔ loads subagent-only extension paths only through child process extension args (0.083375ms)
+  ✔ authorizes child fanout only from exact declared builtin subagent (0.186458ms)
+  ✔ clears all fanout routing env values for non-fanout children (0.079791ms)
+  ✔ inherits routing env only for authorized fanout children (0.128292ms)
+  ✔ prefers the current subagent run id over inherited ancestor ids for nested fanout routing (0.08675ms)
+  ✔ does not let direct MCP tools authorize child fanout (1.116625ms)
+  ✔ keeps child-safe fanout registration in explicit extensions mode (0.072709ms)
+  ✔ emits an empty prompt file when replace mode is used with an empty prompt (0.191166ms)
+✔ buildPiArgs system prompt mode wiring (34.022708ms)
+▶ DM_CODING_AGENT_DIR runtime paths
+  ✔ resolves the agent dir dynamically and loads extension config from it (6.193ms)
+  ✔ discovers user agents, chains, and settings under the configured agent dir (169.572917ms)
+  ✔ resolves user skills, settings skills, and package skills from the configured agent dir (114.17075ms)
+  ✔ records run history and cleans session artifacts under the configured agent dir (3.42075ms)
+  ✔ uses the configured agent dir for subagent bridge instruction files (1.168708ms)
+✔ DM_CODING_AGENT_DIR runtime paths (303.725792ms)
+▶ getPiSpawnCommand
+  ✔ honors explicit DM_SUBAGENT_PI_BINARY override on any platform (0.543417ms)
+  ✔ ignores a blank DM_SUBAGENT_PI_BINARY override (0.053125ms)
+  ✔ uses plain dm on non-Windows even when argv1 is a runnable JS file (0.75625ms)
+  ✔ uses plain dm on non-Windows even when the CLI script can be resolved from package bin (0.241167ms)
+  ✔ falls back to plain dm command on non-Windows when CLI script cannot be resolved (0.162ms)
+  ✔ uses node + argv1 script on Windows when argv1 is runnable JS (0.308542ms)
+  ✔ resolves CLI script from package bin when argv1 is not runnable JS (0.230333ms)
+  ✔ falls back to dm when Windows CLI script cannot be resolved (2.870625ms)
+  ✔ walks from package main entry to resolve package bin (4.379792ms)
+✔ getPiSpawnCommand (10.231542ms)
+▶ getPiSpawnCommand with piPackageRoot
+  ✔ resolves CLI script via piPackageRoot when argv1 is not runnable (0.303125ms)
+✔ getPiSpawnCommand with piPackageRoot (0.468417ms)
+▶ resolveWindowsPiCliScript
+  ✔ supports package bin as string (0.16525ms)
+✔ resolveWindowsPiCliScript (0.231542ms)
+▶ buildDoctorReport
+  ✔ formats a bounded successful environment summary (4.014209ms)
+  ✔ keeps reporting when a directory or discovery check fails (0.569ms)
+✔ buildDoctorReport (4.9565ms)
+▶ dynamic fanout helpers
+  ✔ resolves JSON Pointers and materializes item templates (1.860375ms)
+  ✔ rejects missing structured sources, over-limit arrays, duplicate keys, colliding ids, and bad templates (0.377333ms)
+  ✔ allows config maxItems defaults and handles empty arrays deterministically (0.3075ms)
+  ✔ rejects malformed dynamic-like shapes before they can run as static parallel (0.12575ms)
+  ✔ accepts a runner-injected parentSessionId on the parallel template but keeps it out of user-facing validation (0.093542ms)
+  ✔ accepts toolBudget on dynamic parallel templates (0.07325ms)
+  ✔ validates source ordering and collect name collisions (0.07325ms)
+  ✔ validates appended steps against prior and reserved output names (0.097792ms)
+  ✔ collects ordered child result records and validates aggregate schema (6.181042ms)
+✔ dynamic fanout helpers (9.697041ms)
+▶ DM_SUBAGENT_EXTRA_AGENT_DIRS discovery
+  ✔ discovers agents from the env-provided dirs as a 'user' source (156.213416ms)
+  ✔ scans every directory listed (PATH-style delimiter) (6.839042ms)
+  ✔ lets a local user agent override a bundled one of the same name (4.25ms)
+  ✔ ignores the env var when unset or empty (5.552833ms)
+✔ DM_SUBAGENT_EXTRA_AGENT_DIRS discovery (173.591125ms)
+▶ createFileCoalescer
+  ✔ coalesces duplicate schedule calls per file (0.569208ms)
+  ✔ allows different files to schedule independently (0.077542ms)
+  ✔ clear cancels all pending handlers (0.069458ms)
+✔ createFileCoalescer (1.079584ms)
+▶ foreground tool-call compaction
+  ✔ stores compact tool-call summaries instead of raw message payloads (0.576083ms)
+  ✔ keeps expanded generic tool-call previews bounded (0.082959ms)
+  ✔ does not keep an empty toolCalls array after compaction (0.046083ms)
+  ✔ formats array-based web search previews clearly (0.111333ms)
+  ✔ formats fetch_content urls clearly (0.040375ms)
+  ✔ falls back to generic array previews (0.039ms)
+✔ foreground tool-call compaction (1.292791ms)
+▶ resolveSubagentContext
+  ✔ defaults to fresh (0.241333ms)
+  ✔ accepts fork (0.044209ms)
+✔ resolveSubagentContext (0.631958ms)
+▶ createForkContextResolver
+  ✔ fresh mode never calls createBranchedSession (0.182417ms)
+  ✔ fails fast when parent session file is missing (0.161ms)
+  ✔ fails fast when leaf id is missing (0.058792ms)
+  ✔ opens a throwaway manager from the persisted parent session file (1.366416ms)
+  ✔ creates forked sessions through the default package opener (6.285208ms)
+  ✔ creates parallel forks from a live parent without moving its state or retaining child locks (2.111208ms)
+  ✔ fails clearly for an unflushed user-only parent (0.596875ms)
+  ✔ creates isolated branched sessions per index (parallel and chain compatible) (0.967792ms)
+  ✔ memoizes per index to keep behavior deterministic (0.42175ms)
+  ✔ persists a forked session when DM defers writing the branched file (0.54675ms)
+  ✔ removes signed Anthropic thinking blocks before forwarding a forked session (0.607833ms)
+  ✔ allows unsigned thinking blocks in forked sessions (0.428333ms)
+  ✔ allows non-Anthropic signed thinking blocks in forked sessions (0.421334ms)
+  ✔ removes redacted thinking blocks before fallback-persisting a forked session (0.458083ms)
+  ✔ fails clearly when branch extraction returns a missing child file without fallback state (0.330416ms)
+  ✔ does not silently fallback to fresh when branch extraction fails (0.287ms)
+✔ createForkContextResolver (15.480667ms)
+▶ getFinalOutput
+  ✔ uses the last non-empty text part in the latest assistant message (0.565417ms)
+  ✔ prefers final text over progress text in a multi-part assistant message (0.162125ms)
+  ✔ falls back to an older assistant message when the latest text is whitespace-only (0.057958ms)
+  ✔ falls back to an older assistant message when the latest assistant message is tool-only (0.053333ms)
+  ✔ prefers an earlier explicit acceptance report over later summary-only text (0.073375ms)
+  ✔ prefers an earlier json-fenced acceptance report over later summary-only text (0.0995ms)
+  ✔ does not prefer provider-error acceptance reports (0.048208ms)
+  ✔ returns empty output when all assistant text is empty or whitespace-only (0.053541ms)
+  ✔ does not use provider-error assistant text as fallback output (0.055042ms)
+  ✔ preserves surrounding whitespace on the selected non-empty text (0.070959ms)
+✔ getFinalOutput (1.703959ms)
+▶ subagent extension child mode
+  ✔ collapses tool detail before direct subagent tool execution (2322.407167ms)
+  ✔ does not show async badge for explicit foreground clarify chain calls (868.68575ms)
+  ✔ returns before registering anything for non-fanout children (865.562042ms)
+  ✔ returns before registering anything for fanout children (864.64575ms)
+  ✔ does not double-register the child-safe subagent tool when index and fanout-child both load (843.906125ms)
+  ✔ lets fanout children call read-only list but blocks mutating management actions (849.071541ms)
+✔ subagent extension child mode (6614.892208ms)
+▶ resolveIntercomBridgeMode
+  ✔ defaults unknown values to always (0.255708ms)
+  ✔ accepts explicit modes (0.046ms)
+✔ resolveIntercomBridgeMode (0.635084ms)
+▶ resolveIntercomSessionTarget
+  ✔ prefers an explicit session name (0.073125ms)
+  ✔ uses a runtime-only subagent chat alias when unnamed (0.036792ms)
+✔ resolveIntercomSessionTarget (0.170542ms)
+▶ resolveSubagentIntercomTarget
+  ✔ builds stable child session targets from run metadata (0.110708ms)
+✔ resolveSubagentIntercomTarget (0.151583ms)
+▶ diagnoseIntercomBridge
+  ✔ reports the native supervisor channel as available without external package discovery (0.104916ms)
+  ✔ does not read external intercom config when bridge mode is off (0.052083ms)
+✔ diagnoseIntercomBridge (0.227125ms)
+▶ resolveIntercomBridge
+  ✔ activates when mode/context permit and an orchestrator target exists (0.422542ms)
+  ✔ stays inactive for fresh context when mode is fork-only (0.155916ms)
+  ✔ loads custom instructions from instructionFile (0.653ms)
+  ✔ uses stronger default instructions for fork-aware coordination (0.180958ms)
+✔ resolveIntercomBridge (1.478709ms)
+▶ applyIntercomBridgeToAgent
+  ✔ injects intercom tool and prompt instructions (0.344167ms)
+  ✔ is idempotent (0.046958ms)
+  ✔ does not block native supervisor tools for agents with explicit extension allowlists (0.033167ms)
+✔ applyIntercomBridgeToAgent (0.462ms)
+▶ createJsonlWriter
+  ✔ writes lines with trailing newline (0.5705ms)
+  ✔ pauses on backpressure and resumes on drain (0.096625ms)
+  ✔ closes stream once (0.320291ms)
+  ✔ returns no-op writer when file path is undefined (0.284375ms)
+  ✔ stops writing when maxBytes exceeded without pausing source (0.153209ms)
+  ✔ allows writes up to exactly maxBytes (0.270041ms)
+✔ createJsonlWriter (2.737542ms)
+▶ live async resume interrupt
+  ✔ interrupts a resolved live async child before the caller sends a follow-up (3.23175ms)
+  ✔ still interrupts a live async child when the OS signal is unavailable (ENOSYS on Windows) (8.591917ms)
+  ✔ does not report success or leave a stale request when the runner pid is gone (4.707583ms)
+✔ live async resume interrupt (17.592375ms)
+▶ model fallback helpers
+  ✔ keeps explicit provider/model ids unchanged (0.35625ms)
+  ✔ resolves a bare id when there is exactly one registry match (0.055666ms)
+  ✔ preserves thinking suffix when resolving a bare id (0.2495ms)
+  ✔ leaves ambiguous bare ids untouched (0.052ms)
+  ✔ prefers the current provider when an ambiguous bare id exists there (0.044417ms)
+  ✔ falls back to the unique registry match when the current provider does not offer the model (0.037166ms)
+  ✔ builds a deduplicated ordered candidate list (0.318ms)
+  ✔ applies the current provider preference to fallback candidates too (0.051959ms)
+  ✔ detects retryable provider/model failures (0.500208ms)
+  ✔ does not treat ordinary task/tool failures as retryable model failures (0.098667ms)
+✔ model fallback helpers (2.231333ms)
+▶ resolveSubagentModelOverride (cross-session inherit, issue #266)
+  ✔ inherits the parent session model when no model is requested (0.08975ms)
+  ✔ inherits the parent session model when the model is the "inherit" sentinel (0.033208ms)
+  ✔ inherits the parent session model when the agent config sets model: false (delegate) (0.022625ms)
+  ✔ treats an empty or whitespace-only model as inherit (0.027458ms)
+  ✔ trims surrounding whitespace from the "inherit" sentinel (0.024ms)
+  ✔ keeps an explicit provider/id model unchanged (0.02ms)
+  ✔ resolves an explicit bare id against the registry, not the parent (0.019458ms)
+  ✔ returns undefined when inheriting but no parent model is known (0.027709ms)
+  ✔ never emits the literal "inherit" string as a model (0.039042ms)
+✔ resolveSubagentModelOverride (cross-session inherit, issue #266) (0.391375ms)
+▶ fuzzyResolveModel / normalizeModelSegment
+  ✔ normalizes dots, underscores, case, and repeated dashes (0.058083ms)
+  ✔ fuzzy-matches a bare id with separator/case differences (0.037375ms)
+  ✔ fuzzy-matches a bare id with an optional trailing date stamp (0.046459ms)
+  ✔ does not strip arbitrary trailing 8-digit numbers as date stamps (0.024916ms)
+  ✔ fuzzy-matches an undated query against a dated registry id (0.031875ms)
+  ✔ fuzzy-matches a qualified provider/id with case/separator differences (0.038875ms)
+  ✔ does not switch providers for a qualified query (0.023875ms)
+  ✔ prefers the current provider for an ambiguous bare fuzzy id (0.031291ms)
+  ✔ returns undefined for an ambiguous bare fuzzy id with no preferred provider (0.01875ms)
+  ✔ returns undefined when nothing fuzzy-matches (0.021417ms)
+✔ fuzzyResolveModel / normalizeModelSegment (0.41075ms)
+▶ resolveModelCandidate fuzzy fallback
+  ✔ resolves a bare id with case/separator differences via fuzzy fallback (0.042ms)
+  ✔ resolves a bare id with a trailing date stamp via fuzzy fallback (0.020125ms)
+  ✔ resolves a qualified provider/id with case differences via fuzzy fallback (0.021583ms)
+  ✔ preserves the thinking suffix through fuzzy resolution (0.037208ms)
+  ✔ still prefers exact registry matches over fuzzy (0.019209ms)
+  ✔ leaves an unknown qualified model unchanged instead of switching providers (0.025333ms)
+  ✔ leaves an unknown bare id unchanged when no fuzzy match exists (0.021417ms)
+✔ resolveModelCandidate fuzzy fallback (0.24ms)
+▶ resolveSubagentModelOverride scope enforcement
+  ✔ is a no-op when scope is not enforced (0.043ms)
+  ✔ throws for an explicit out-of-scope model (0.237416ms)
+  ✔ warns (and still returns the model) for an inherited out-of-scope model (0.090708ms)
+  ✔ warns for an inherited parent-session model that is out of scope (0.033042ms)
+  ✔ passes through an in-scope explicit model without warning or error (0.02775ms)
+  ✔ checks the resolved (canonicalized) model against the scope (0.0275ms)
+  ✔ ignores a thinking suffix when checking scope (0.038041ms)
+  ✔ warns for out-of-scope fallback models while keeping them available (0.0485ms)
+✔ resolveSubagentModelOverride scope enforcement (0.609334ms)
+▶ model info helpers
+  ✔ does not choose arbitrary metadata for ambiguous bare model ids (0.678083ms)
+  ✔ uses the preferred provider for ambiguous bare model metadata (0.135ms)
+  ✔ matches provider-qualified model metadata before bare ids (0.109041ms)
+  ✔ keeps the legacy full thinking list for reasoning models without per-level metadata (0.715709ms)
+  ✔ keeps the legacy full thinking list when older model metadata omits reasoning (0.137792ms)
+  ✔ filters levels only when per-level metadata is present (0.225084ms)
+  ✔ honors metadata that marks off unsupported (0.202084ms)
+✔ model info helpers (3.548833ms)
+▶ subagents.modelScope discovery
+  ✔ exposes a user modelScope from discoverAgents (126.958417ms)
+  ✔ returns undefined when no modelScope is configured (6.203167ms)
+  ✔ prefers project modelScope over user modelScope (8.400208ms)
+  ✔ falls back to user modelScope when project does not set one (4.277375ms)
+  ✔ rejects an invalid modelScope config at discovery time (2.23675ms)
+  ✔ rejects a malformed modelScope value at discovery time (2.195625ms)
+✔ subagents.modelScope discovery (150.770458ms)
+▶ matchesScopePattern
+  ✔ matches an exact provider/id (0.363917ms)
+  ✔ matches a provider wildcard (0.078875ms)
+  ✔ matches a model-prefix wildcard (0.078875ms)
+  ✔ does not match across providers (0.978ms)
+  ✔ ignores case on both sides (0.095583ms)
+  ✔ strips a thinking suffix before matching (0.089042ms)
+  ✔ does not treat arbitrary colon text as a thinking suffix (0.055917ms)
+  ✔ requires a full match, not a substring (0.059417ms)
+✔ matchesScopePattern (2.236166ms)
+▶ checkModelScope
+  ✔ returns undefined when enforcement is off (3.462042ms)
+  ✔ returns undefined when no allow list is configured (0.390209ms)
+  ✔ returns undefined when the model is in scope (0.340042ms)
+  ✔ returns an error violation for an explicit out-of-scope model (0.700916ms)
+  ✔ returns a warn violation for an inherited out-of-scope model (0.419458ms)
+  ✔ defaults to inherited (warn) severity when source is omitted-ish via inherited (0.048709ms)
+  ✔ strips the thinking suffix from the reported model (0.033125ms)
+  ✔ returns undefined for an undefined model (0.02825ms)
+✔ checkModelScope (5.77975ms)
+▶ parseModelScopeConfig
+  ✔ returns undefined when the field is absent (0.183708ms)
+  ✔ parses a well-formed config (0.072667ms)
+  ✔ parses enforce alone (0.193417ms)
+  ✔ trims allow patterns and drops empties (0.123083ms)
+  ✔ rejects a non-object value (0.6145ms)
+  ✔ rejects a non-boolean enforce (0.341583ms)
+  ✔ rejects a non-array allow (0.201292ms)
+  ✔ rejects an allow array with non-string entries (0.596209ms)
+  ✔ rejects enforce without a non-empty allow list (0.121417ms)
+✔ parseModelScopeConfig (2.686208ms)
+▶ native supervisor channel
+  ✔ delivers requests only to the exact current session id (22.073333ms)
+  ✔ matches supervisor requests against the runtime session id instead of persisted session file path (12.917333ms)
+  ✔ keeps an installed intercom tool and still exposes a native supervisor reply path (11.6705ms)
+✔ native supervisor channel (47.233458ms)
+▶ nested control routing
+  ✔ routes interrupt to an explicit nested id through the control inbox (64.281791ms)
+  ✔ renders nested children in foreground status output (4.507708ms)
+  ✔ scopes child-safe nested status lookup to the inherited route and child address (3.954625ms)
+  ✔ requires an id for child-safe status instead of listing unrelated top-level async runs (0.544958ms)
+  ✔ does not let bare interrupt target hidden nested descendants (1.058709ms)
+  ✔ times out owner-gone nested control and ignores late results (1045.636875ms)
+  ✔ routes resume for live nested runs through the control inbox (52.819875ms)
+  ✔ validates terminal nested resume session files before revive (1.523917ms)
+  ✔ rejects terminal nested resume session files outside trusted roots (1.463625ms)
+  ✔ rejects terminal nested resume session files from sibling run directories (1.556ms)
+  ✔ emits a failed completed nested event when foreground execution throws after start (3.03925ms)
+  ✔ keeps the fanout child control listener alive after control inbox polling errors (430.101ms)
+  ✔ keeps fanout child control requests when result writing fails and retries after recovery (421.601708ms)
+  ✔ negatively acknowledges ownerless fanout child control requests and removes them (213.167666ms)
+✔ nested control routing (2246.223292ms)
+▶ nested route index
+  ✔ indexes routes by root run id in a single directory scan (15.723792ms)
+  ✔ keeps at most one route when a root run id has duplicate route dirs (10.51125ms)
+✔ nested route index (26.877875ms)
+▶ nested event route validation
+  ✔ resolves nested parent addresses with full inherited path (0.54725ms)
+  ✔ ignores unsafe nested parent ids from env (0.08975ms)
+  ✔ resolves only matching contained routes from env (0.821459ms)
+✔ nested event route validation (1.542958ms)
+▶ nested event parsing and projection
+  ✔ projects started, updated, and completed records into async and foreground parent state (2.453417ms)
+  ✔ attaches root children to visible step slices by original step index (1.011958ms)
+  ✔ ignores corrupt, partial, wrong-token, duplicate, and stale records while preserving terminal state (1.64975ms)
+  ✔ detects live descendants attached to terminal step children (0.088416ms)
+  ✔ accepts only complete numeric token usage at the nested event boundary (2.284542ms)
+  ✔ parses only complete jsonl records (0.602125ms)
+✔ nested event parsing and projection (8.23025ms)
+▶ registerSubagentNotify
+  ✔ uses a fallback summary when a background completion is empty (0.795042ms)
+  ✔ preserves non-empty completion summaries (0.10375ms)
+  ✔ preserves session paths in notification content (0.082125ms)
+  ✔ labels paused completions as paused even without an exit code (0.314667ms)
+  ✔ ignores completions for other or missing session ids (0.233417ms)
+  ✔ emits failed completions immediately even while successes are held (0.84175ms)
+  ✔ groups sibling successes into a single notification after the debounce window (0.635ms)
+  ✔ ignores successes from other sessions instead of grouping them (0.932708ms)
+  ✔ does not let another session failure flush held successes (0.146375ms)
+✔ registerSubagentNotify (4.679959ms)
+▶ completion formatting helpers
+  ✔ formatSingleCompletion mirrors the in-handler single message shape (0.0955ms)
+  ✔ formatGroupedCompletion lists each agent with its summary and session (0.043542ms)
+  ✔ buildCompletionDetails derives paused status from state and summary (0.042ms)
+  ✔ buildCompletionDetails falls back to the unknown agent label (0.032083ms)
+✔ completion formatting helpers (0.29275ms)
+✔ direct @earendil-works runtime imports are declared for CI installs (11.008834ms)
+✔ direct dependency declarations are exact version pins (0.22275ms)
+✔ old DM package scope is not used by source or tests (7.195167ms)
+✔ DM package resolution stays export-map safe (10.602833ms)
+▶ isParallelGroup
+  ✔ returns true for a parallel step group (1.383584ms)
+  ✔ returns false for a sequential step (1.033834ms)
+  ✔ returns false when parallel is not an array (0.478084ms)
+✔ isParallelGroup (3.364083ms)
+▶ flattenSteps
+  ✔ returns sequential steps unchanged (0.106208ms)
+  ✔ expands parallel groups into individual steps (0.317125ms)
+  ✔ handles empty steps array (0.043209ms)
+  ✔ handles empty parallel group (0.209709ms)
+✔ flattenSteps (0.827417ms)
+▶ mapConcurrent
+  ✔ processes all items and preserves order (0.513583ms)
+  ✔ respects concurrency limit (33.567083ms)
+  ✔ handles empty input (0.254708ms)
+  ✔ clamps limit=0 to 1 (sequential execution) (35.349584ms)
+  ✔ clamps limit=-1 to 1 (sequential execution) (36.117958ms)
+  ✔ does not stagger by default (10.712875ms)
+  ✔ respects a shared global semaphore across simultaneous calls (32.72025ms)
+  ✔ clamps invalid global semaphore limits to 1 (38.837792ms)
+✔ mapConcurrent (188.740625ms)
+▶ DEFAULT_GLOBAL_CONCURRENCY_LIMIT
+  ✔ is 20 (0.06075ms)
+✔ DEFAULT_GLOBAL_CONCURRENCY_LIMIT (0.096417ms)
+▶ aggregateParallelOutputs
+  ✔ aggregates successful outputs with headers (0.129625ms)
+  ✔ marks failed tasks (0.032583ms)
+  ✔ marks empty output (0.030208ms)
+  ✔ treats whitespace-only output as empty (0.02375ms)
+  ✔ marks skipped tasks (exitCode=-1) distinctly from failed (0.026083ms)
+✔ aggregateParallelOutputs (0.290834ms)
+▶ MAX_PARALLEL_CONCURRENCY
+  ✔ is 4 (0.021666ms)
+✔ MAX_PARALLEL_CONCURRENCY (0.03775ms)
+▶ path.isAbsolute vs startsWith('/')
+  ✔ startsWith('/') misses Windows absolute paths (0.259917ms)
+  ✔ path.isAbsolute is the correct cross-platform check (0.054ms)
+✔ path.isAbsolute vs startsWith('/') (0.659666ms)
+▶ path.join vs template string concatenation
+  ✔ template concatenation produces forward slashes regardless of platform (0.068875ms)
+  ✔ resolveChainPath pattern should use path.join for relative paths (0.042333ms)
+  ✔ parallel subdir naming should use path.join (0.050375ms)
+✔ path.join vs template string concatenation (0.24175ms)
+▶ Path resolution for .agents and ~/.agents
+  ✔ should resolve skills in .agents/skills (144.209125ms)
+  ✔ should resolve skills in ~/.agents/skills (3.339959ms)
+  ✔ should resolve project agents from both .agents and .dm/agents (7.975625ms)
+  ✔ should resolve agents in ~/.agents (124.311708ms)
+✔ Path resolution for .agents and ~/.agents (280.265167ms)
+▶ proactive skill subagent recommendations
+  ✔ recommends available skills referenced by multiple enabled configs (6.542875ms)
+  ✔ filters unavailable orchestration skills and honors config bounds (0.1315ms)
+  ✔ can be disabled and formats guardrails for visible suggestions (0.123834ms)
+  ✔ does not discover skills when disabled and treats discovery failures as no suggestions (0.092084ms)
+✔ proactive skill subagent recommendations (7.278792ms)
+▶ profiles helpers
+  ✔ lists no profiles when the directory is empty (1.627083ms)
+  ✔ applies a saved profile by replacing only settings.subagents (5.743291ms)
+  ✔ rejects profile and provider path traversal names (1.427709ms)
+  ✔ refreshes a provider model catalog and writes a cache file (6.98975ms)
+  ✔ does not count heuristic fallback when official metadata is present (3.556541ms)
+  ✔ reuses a fresh provider model catalog (1.161625ms)
+  ✔ generates quota and quality profiles from sorted provider models (1.636875ms)
+  ✔ does not treat provider names like MiniMax as mini-tier tokens and prefers M3 over M2.7-highspeed for quality (1.72875ms)
+  ✔ checks a profile against the registry and live probe (1.162792ms)
+  ✔ checks short model ids and thinking suffixes against the registry (0.847041ms)
+✔ profiles helpers (26.452667ms)
+▶ prompt-template delegation bridge
+  ✔ emits started/update/response on successful request (0.980292ms)
+  ✔ rebuilds compact tool-call summaries into delegated response messages (0.192125ms)
+  ✔ filters malformed recent output entries in updates (1.978166ms)
+  ✔ returns structured error when no active context (0.162375ms)
+  ✔ accepts requests when delegated cwd differs from active context (0.10725ms)
+  ✔ applies pending cancel when cancel arrives before request (0.094208ms)
+  ✔ cancels in-flight delegated execution (0.138125ms)
+  ✔ accepts tasks payloads and emits parallelResults (0.149667ms)
+  ✔ marks missing parallel task results as errors (0.95725ms)
+✔ prompt-template delegation bridge (5.304459ms)
+▶ prompt workflows
+  ✔ discovers project workflows over user workflows (16.304958ms)
+  ✔ runs a named workflow through native subagent execution (1.895083ms)
+  ✔ runs prompt templates as a native chain (1.880958ms)
+✔ prompt workflows (20.792708ms)
+▶ DEFAULT_SUBAGENT_MAX_DEPTH
+  ✔ is 2 (0.718125ms)
+✔ DEFAULT_SUBAGENT_MAX_DEPTH (1.4895ms)
+▶ DEFAULT_MAX_SUBAGENT_SPAWNS_PER_SESSION
+  ✔ is 40 (0.927792ms)
+✔ DEFAULT_MAX_SUBAGENT_SPAWNS_PER_SESSION (2.327541ms)
+▶ normalizeMaxSubagentDepth
+  ✔ accepts integers >= 0 (0.191041ms)
+  ✔ rejects negatives and non-integers (0.107084ms)
+✔ normalizeMaxSubagentDepth (0.38275ms)
+▶ normalizeMaxSubagentSpawnsPerSession
+  ✔ accepts integers >= 0 (0.087292ms)
+  ✔ rejects negatives and non-integers (0.079167ms)
+✔ normalizeMaxSubagentSpawnsPerSession (0.233125ms)
+▶ resolveMaxSubagentSpawnsPerSession
+  ✔ uses env when present (0.073875ms)
+  ✔ falls back to config when env is absent (0.060958ms)
+  ✔ falls back to default when neither env nor config is valid (0.0735ms)
+✔ resolveMaxSubagentSpawnsPerSession (0.266291ms)
+▶ resolveCurrentMaxSubagentDepth
+  ✔ uses env when present (0.095333ms)
+  ✔ falls back to config when env is absent (0.058667ms)
+  ✔ falls back to default when neither env nor config is valid (0.049458ms)
+✔ resolveCurrentMaxSubagentDepth (0.238542ms)
+▶ top-level parallel config helpers
+  ✔ resolves maxTasks from config or falls back to the default (0.060791ms)
+  ✔ resolves concurrency from per-call override, config, or default (0.049291ms)
+✔ top-level parallel config helpers (0.144416ms)
+▶ resolveChildMaxSubagentDepth
+  ✔ keeps the inherited max when agent override is absent (0.061459ms)
+  ✔ tightens to the lower per-agent max (0.039542ms)
+  ✔ does not relax an already stricter inherited max (0.041125ms)
+✔ resolveChildMaxSubagentDepth (0.179166ms)
+▶ checkSubagentDepth
+  ✔ not blocked at depth=0, max=2 (0.092583ms)
+  ✔ uses config max depth when env is absent (0.042666ms)
+  ✔ not blocked at depth=1, max=2 (0.995125ms)
+  ✔ blocks at depth=1, max=1 after one nested level (0.193333ms)
+  ✔ blocked at depth=2, max=2 (0.129083ms)
+  ✔ blocked at depth=3, max=2 (0.262792ms)
+  ✔ blocked at depth=0, max=0 (disables subagent entirely) (0.083291ms)
+  ✔ defaults to depth=0, max=2 when env vars unset (0.056958ms)
+  ✔ not blocked when depth is invalid (NaN) (0.048833ms)
+✔ checkSubagentDepth (2.063042ms)
+▶ getSubagentDepthEnv
+  ✔ increments from depth=0 (0.082583ms)
+  ✔ increments from depth=1 (0.039166ms)
+  ✔ uses provided max depth override (0.042667ms)
+  ✔ respects inherited DM_SUBAGENT_MAX_DEPTH when override is absent (10.2965ms)
+  ✔ uses the explicit child override even when a looser inherited env max exists (0.205333ms)
+  ✔ falls back to depth=1 when env var is invalid (NaN) (0.459625ms)
+✔ getSubagentDepthEnv (11.270792ms)
+✔ row clips content to the available width (1.208666ms)
+✔ row normalizes multiline content before clipping (0.1435ms)
+✔ row keeps styled multiline content within the available width (0.104625ms)
+✔ compact chain rendering uses workflow graph spans for dynamic fanout results (15.171292ms)
+✔ compact chain rendering shows failed zero-child dynamic fanout groups (0.293375ms)
+✔ expanded chain rendering uses workflow graph spans for dynamic fanout results (0.241291ms)
+✔ compact multi-result rendering shows total cost in the header (0.160875ms)
+✔ static sequential and static parallel chain rendering keep existing labels (0.206584ms)
+▶ result intercom formatter
+  ✔ builds one grouped intercom payload with status counts and child sections (0.542167ms)
+  ✔ advertises async revive only for single-child results with an existing session (0.410958ms)
+  ✔ advertises indexed revive for multi-child async results with existing child sessions (0.355416ms)
+  ✔ does not advertise async revive for missing child session files (0.069167ms)
+  ✔ attaches compact nested children under their parent result child without route secrets (0.273375ms)
+  ✔ keeps full child summaries inside grouped payloads (0.285458ms)
+  ✔ formats compact grouped receipts with artifacts and sessions (0.126583ms)
+  ✔ strips heavy output fields from receipt details (0.062709ms)
+  ✔ resolves paused and detached statuses (0.057458ms)
+✔ result intercom formatter (2.625917ms)
+▶ subagent extension RPC bridge
+  ✔ emits ready and answers ping with versioned capability metadata (0.622791ms)
+  ✔ replies to malformed request ids on the safe unknown channel (0.158333ms)
+  ✔ delegates status through the existing executor action (0.441333ms)
+  ✔ forces spawn requests onto the existing async execution path (0.134416ms)
+  ✔ rejects foreground or management spawn requests before executor dispatch (0.176959ms)
+  ✔ delegates interrupt through the existing executor action (0.095709ms)
+  ✔ uses the existing async timeout control path for stop (2.154959ms)
+  ✔ rejects stop requests for async runs from a different session (0.75725ms)
+✔ subagent extension RPC bridge (5.071334ms)
+▶ subagent run id resolver
+  ✔ prefers exact foreground, then exact async, then exact nested before prefix matches (38.978291ms)
+  ✔ reports one combined ambiguity for prefixes across namespaces (70.770625ms)
+  ✔ limits nested lookup to active state routes when state is provided (25.610042ms)
+  ✔ limits nested lookup to descendants of a scoped child address (2.198916ms)
+  ✔ reports async prefix ambiguity without parsing resolver error text (57.59825ms)
+  ✔ rejects unsafe nested id tokens before lookup (0.138625ms)
+✔ subagent run id resolver (195.970708ms)
+▶ async run status inspection
+  ✔ repairs stale running status and reports diagnosis plus result path (7.402958ms)
+  ✔ shows parallel mode and aggregate progress for top-level async parallel runs (1.621083ms)
+  ✔ tails a readable transcript from async output artifacts (3.023166ms)
+  ✔ does not fall back to another child output when an explicit transcript index output is missing (1.655709ms)
+  ✔ refuses to tail status outputFile paths outside the async directory (0.850042ms)
+  ✔ refuses symlink session transcript paths even under trusted roots (2.128833ms)
+  ✔ shows an active read-only fleet view with transcript commands (11.070834ms)
+  ✔ scopes fleet active-run discovery to the current session (11.1765ms)
+  ✔ refuses transcript reads for async runs owned by another session (0.703334ms)
+  ✔ does not fall back to aggregate result output for an explicit completed child index (0.779ms)
+  ✔ surfaces steering counts and timestamps in exact and list status (10.776334ms)
+  ✔ shows nested runs under owning steps with exact status hints (5.065375ms)
+  ✔ repairs stale nested async descendants before rendering root status (8.347583ms)
+  ✔ shows a warning when nested projection fails for detailed status (2.406875ms)
+  ✔ shows a warning when nested projection fails for active status lists (9.518583ms)
+  ✔ resolves exact nested run ids from the nested registry (43.1705ms)
+  ✔ shows indexed revive guidance for completed multi-child async runs with child sessions (1.028ms)
+  ✔ uses original child indexes when result metadata contains invalid children (0.437041ms)
+  ✔ labels chain parallel group children with logical step and agent numbers (0.86825ms)
+  ✔ shows expected intercom target for still-running async steps (0.927541ms)
+  ✔ rejects ambiguous async run id prefixes (50.832833ms)
+  ✔ rejects path-like async run ids (0.252541ms)
+  ✔ does not advertise revive for result fallback with only a top-level session file (0.784416ms)
+  ✔ treats a top-level completed result as one transcript child (2.862416ms)
+  ✔ validates completed result transcript indexes as integers (1.010333ms)
+  ✔ falls back to an existing result when async dir has no status file (0.68175ms)
+✔ async run status inspection (180.53525ms)
+▶ scheduled-runs helpers
+  ✔ isScheduledRunAction narrows the four actions (0.282834ms)
+  ✔ scheduledRunsEnabled is true only when explicitly enabled (0.055667ms)
+  ✔ scheduledRunStorePath is stable per cwd+session and distinct otherwise (0.470333ms)
+  ✔ parseScheduledRunTime handles relative delays and ISO timestamps (0.292875ms)
+  ✔ parseScheduledRunTime rejects past, zero, malformed, and ambiguous schedules (1.630917ms)
+✔ scheduled-runs helpers (3.209667ms)
+▶ ScheduledRunManager create/list/status/cancel
+  ✔ rejects schedule actions when the feature is disabled (0.934708ms)
+  ✔ creates a scheduled job and arms a timer (2.442958ms)
+  ✔ requires exactly one execution mode (0.606125ms)
+  ✔ requires a schedule and rejects fork/async-false/clarify-true (0.206542ms)
+  ✔ rejects a past schedule time (0.196875ms)
+  ✔ enforces maxPending (0.551042ms)
+  ✔ status resolves by exact id and prefix, and rejects ambiguous/missing (0.642875ms)
+  ✔ cancel clears the timer and marks the job canceled (0.806917ms)
+  ✔ cancel refuses a terminal job (0.59525ms)
+  ✔ reports malformed persisted job records instead of dropping them (0.241709ms)
+  ✔ reports JSON parse errors from a corrupted persisted store (0.236917ms)
+✔ ScheduledRunManager create/list/status/cancel (9.028042ms)
+▶ ScheduledRunManager firing
+  ✔ launches the sanitized async run when the timer fires (1.854667ms)
+  ✔ marks the job failed when launch returns an error result (0.760209ms)
+  ✔ marks the job failed when launch throws (0.7335ms)
+  ✔ re-arms instead of launching when a capped timer fires before runAt (0.712ms)
+  ✔ cancel before fire prevents the launch (0.45325ms)
+✔ ScheduledRunManager firing (5.139083ms)
+▶ ScheduledRunManager restart restore
+  ✔ re-arms pending jobs and marks stale jobs missed after a restart (0.912333ms)
+  ✔ store path lives under the scheduled runs dir by default (0.031792ms)
+✔ ScheduledRunManager restart restore (1.088084ms)
+▶ SubagentParams schema
+  ✔ includes context field for fresh/fork execution mode (0.605125ms)
+  ✔ includes count and concurrency on top-level parallel mode (0.118209ms)
+  ✔ allows runtime validation of management and control action strings (0.060083ms)
+  ✔ includes foreground timeout aliases and turn budget (0.107666ms)
+  ✔ includes subagent control fields (0.166041ms)
+  ✔ does not emit description-only schema nodes (0.258917ms)
+  ✔ does not emit array-typed schema nodes without items (0.229625ms)
+  ✔ keeps only top-level parameter descriptions to keep the provider payload compact (0.371291ms)
+  ✔ preserves TypeBox metadata while pruning provider-visible descriptions (0.076708ms)
+  ✔ does not emit provider-rejected schema shapes (0.338417ms)
+  ✔ uses provider-friendly anyOf unions for flexible fields and chain items (0.132167ms)
+  ✔ validates representative flexible field values with TypeBox compiler (9.941083ms)
+✔ SubagentParams schema (12.885416ms)
+▶ normalizeSingleOutputOverride
+  ✔ treats boolean and string false as disabled output (0.753ms)
+  ✔ treats boolean and string true as the configured default output (0.113125ms)
+  ✔ passes explicit non-empty output paths through (0.070042ms)
+✔ normalizeSingleOutputOverride (1.704333ms)
+▶ resolveSingleOutputPath
+  ✔ does not resolve disabled or boolean-like output values (0.117166ms)
+  ✔ keeps absolute paths unchanged (0.116666ms)
+  ✔ resolves relative paths against requested cwd (0.078666ms)
+  ✔ resolves relative paths against runtime cwd when requested cwd is absent (0.051ms)
+  ✔ resolves relative requested cwd from runtime cwd before resolving output (0.064834ms)
+  ✔ resolves relative output paths against an explicit artifact base (0.060542ms)
+✔ resolveSingleOutputPath (0.614458ms)
+▶ injectSingleOutputInstruction
+  ✔ appends output instruction with resolved path (0.153709ms)
+✔ injectSingleOutputInstruction (0.18125ms)
+▶ injectOutputPathSystemPrompt
+  ✔ adds the authoritative runtime output path to the system prompt (0.071334ms)
+  ✔ leaves prompts unchanged when no output path is active (0.045292ms)
+✔ injectOutputPathSystemPrompt (0.148667ms)
+▶ resolveSingleOutput
+  ✔ keeps agent-written file content when the file changed during the run (0.594334ms)
+  ✔ falls back to persisting the assistant output when the file was not changed (0.473791ms)
+  ✔ preserves read errors from changed output paths (0.423541ms)
+✔ resolveSingleOutput (1.53925ms)
+▶ formatSavedOutputReference
+  ✔ includes absolute path, human-readable size, and line count (0.177ms)
+  ✔ formats larger byte sizes in KB (0.08425ms)
+✔ formatSavedOutputReference (0.291666ms)
+▶ validateFileOnlyOutputMode
+  ✔ requires an output path for file-only mode (0.0515ms)
+✔ validateFileOnlyOutputMode (0.069583ms)
+▶ finalizeSingleOutput
+  ✔ formats saved-path messaging around the already-resolved output (0.082125ms)
+  ✔ returns only the saved-output reference in file-only mode (0.054708ms)
+  ✔ does not add save messaging on failed runs (0.041459ms)
+✔ finalizeSingleOutput (0.208583ms)
+▶ skills filesystem fallback
+  ✔ discovers project skills from filesystem paths (118.816125ms)
+  ✔ discovers project skills nested below grouping directories (10.2925ms)
+  ✔ stops recursive project skill discovery at the first SKILL.md anchor (4.995875ms)
+  ✔ skips hidden directories and node_modules while recursing for project skills (5.360583ms)
+  ✔ keeps direct markdown skills from explicit settings roots after parent recursion (2.747625ms)
+  ✔ keeps nested skills from higher-priority explicit settings roots after parent recursion (2.983667ms)
+  ✔ keeps nested skills from higher-priority explicit settings roots when the root path is duplicated (4.536ms)
+  ✔ resolves and reads skill content via filesystem fallback (3.836917ms)
+  ✔ builds lazy skill references instead of inlining full skill bodies (5.570709ms)
+  ✔ escapes XML-sensitive skill metadata in lazy references (2.846958ms)
+  ✔ does not expose dm-subagents as a child-injectable skill (7.602834ms)
+  ✔ classifies package-provided skills as project-package (3.31ms)
+  ✔ prefers project skills over project-package skills with the same name (4.781166ms)
+  ✔ discovers skills from project settings packages (3.751917ms)
+  ✔ discovers skills from project settings npm package sources (4.236584ms)
+  ✔ discovers skills from the current cwd package (3.052458ms)
+  ✔ falls back to the runtime cwd when the execution cwd lacks the skill (4.834041ms)
+  ✔ discovers skills from user settings packages (132.528958ms)
+  ✔ discovers skills from user settings git package sources (128.096416ms)
+  ✔ discovers skills from user settings scoped npm package sources (117.078583ms)
+  ✔ surfaces malformed project settings files instead of silently ignoring them (1.643209ms)
+  ✔ surfaces malformed explicit settings package manifests instead of silently ignoring them (1.792084ms)
+✔ skills filesystem fallback (575.592041ms)
+▶ slash subagent bridge requester context
+  ✔ uses request ctx instead of stale fallback context when provided (0.409584ms)
+✔ slash subagent bridge requester context (0.729833ms)
+▶ parseSingleTaskToken
+  ✔ parses a quoted task (0.293792ms)
+  ✔ parses an agent with inline config and no task (0.13375ms)
+  ✔ parses a task via -- delimiter (0.044ms)
+  ✔ parses extended metadata in inline config (0.054666ms)
+  ✔ ignores a non-positive count (0.052791ms)
+✔ parseSingleTaskToken (0.833ms)
+▶ parseGroupSegment
+  ✔ parses a static parallel group with two quoted tasks (1.118375ms)
+  ✔ rejects groups with a single task (0.568167ms)
+  ✔ rejects groups with unbalanced parentheses (0.383041ms)
+  ✔ parses a trailing group-options suffix (0.67ms)
+  ✔ defaults to an empty group config without a suffix (0.536125ms)
+  ✔ rejects a non-bracketed group suffix (0.064125ms)
+✔ parseGroupSegment (3.869625ms)
+▶ hasGroupSyntax
+  ✔ detects parentheses (0.083416ms)
+  ✔ does not treat a bare pipe as group syntax (0.027292ms)
+  ✔ ignores parens inside quotes (0.021083ms)
+  ✔ returns false for plain chain input (0.019041ms)
+  ✔ does not treat unquoted parens inside a -- task as group syntax (0.024583ms)
+  ✔ still detects a group that opens a step (0.021708ms)
+✔ hasGroupSyntax (0.259417ms)
+▶ parseChainExpression
+  ✔ parses sequential + group + sequential (0.122291ms)
+  ✔ rejects expression without arrows (0.035ms)
+  ✔ rejects groups with one task (0.038833ms)
+  ✔ respects quotes when splitting on arrows (0.044667ms)
+  ✔ allows balanced parens in a -- task after a group (0.04325ms)
+  ✔ still rejects truly unmatched parens in a non-group segment (0.036084ms)
+✔ parseChainExpression (0.38025ms)
+▶ buildChainExpressionSteps
+  ✔ emits a chain with a { parallel: [...] } step for groups (88.314417ms)
+  ✔ preserves backward-compatible linear chain behavior (1.986208ms)
+  ✔ falls back to shared task when first step has no task (1.952917ms)
+  ✔ reports parallel group errors as notifications (0.122292ms)
+  ✔ propagates inline metadata onto chain steps (1.621375ms)
+  ✔ applies count only inside a parallel group (1.954583ms)
+  ✔ propagates group-level options onto the parallel step (1.558083ms)
+  ✔ rejects an invalid acceptance level (1.673ms)
+  ✔ rejects inline acceptance levels that require object contracts (1.609708ms)
+  ✔ loads an inline outputSchema path and rejects a missing one (3.219916ms)
+  ✔ keeps a bare pipe in a -- task on the legacy single-agent path (1.620417ms)
+  ✔ keeps unquoted parens in a -- task on the legacy single-agent path (1.759459ms)
+  ✔ allows balanced parens in a -- task after a group step (1.546125ms)
+  ✔ exports a stable parallel group usage hint (0.045834ms)
+✔ buildChainExpressionSteps (109.1545ms)
+▶ async stale-run reconciliation
+  ✔ classifies pid liveness without treating EPERM as dead (0.312042ms)
+  ✔ marks a running async run failed when the runner pid is dead and no result exists (1.748834ms)
+  ✔ includes runner stderr diagnostics when repairing a stale startup crash (1.275625ms)
+  ✔ keeps stale repair successful when the event log cannot be appended (1.080833ms)
+  ✔ repairs stale status with per-child result outcomes (0.870541ms)
+  ✔ fails a stale run when a live pid has not updated beyond the stale threshold (0.952917ms)
+  ✔ preserves an existing result instead of overwriting it with stale-run failure (0.934875ms)
+✔ async stale-run reconciliation (7.6ms)
+▶ status format helpers
+  ✔ formats activity labels consistently (0.605ms)
+  ✔ aggregates step status and parallel outcomes (0.142666ms)
+✔ status format helpers (1.34225ms)
+▶ dm-subagents source-layout contract
+  ✔ tracks nicobailon source-layout v0.33.1 as dm-subagents (0.545167ms)
+  ✔ keeps source-layout tools and /agents command visible (0.224625ms)
+  ✔ documents only the capabilities present in the nicobailon source layout (0.7525ms)
+  ✔ does not leak user-facing upstream subagents branding (8.264166ms)
+✔ dm-subagents source-layout contract (10.225584ms)
+▶ subagent control attention state
+  ✔ marks a run as needing attention only after the idle threshold (0.3475ms)
+  ✔ builds compact needs-attention control events (0.345042ms)
+  ✔ supports a specific attention message (0.057166ms)
+  ✔ builds terminal completion guard control events (0.045375ms)
+  ✔ defaults notifications to active-long-running and needs attention (0.06825ms)
+  ✔ defaults active-long-running notices to elapsed time only (0.061625ms)
+  ✔ supports opt-in turn and token long-running thresholds (0.056166ms)
+  ✔ resolves custom notification config (0.071042ms)
+  ✔ falls back to defaults for invalid non-empty notification arrays (0.061584ms)
+  ✔ allows empty notification arrays to disable notifications (0.087584ms)
+  ✔ formats control notices with a proactive hint and concrete commands (0.156708ms)
+  ✔ formats active-long-running notices as informational (0.091917ms)
+  ✔ formats terminal completion guard notices without live-run commands (0.054708ms)
+  ✔ formats intercom notifications with the same control commands (0.057791ms)
+  ✔ dedupes notifications once per child target and attention state (0.090459ms)
+✔ subagent control attention state (2.710834ms)
+▶ subagent prompt runtime
+  ✔ nudges after the tool budget soft limit and blocks configured tools after hard (0.952167ms)
+  ✔ delivers steering inbox requests as mid-run user messages (5.004333ms)
+  ✔ registered structured_output tool accepts valid schema output and writes the capture file (3.182667ms)
+  ✔ strips only the project context block (0.122ms)
+  ✔ strips only the inherited skills block (0.097542ms)
+  ✔ can strip both inherited sections together (0.201625ms)
+  ✔ injects a child-only boundary that forbids proposing or running subagents (0.147583ms)
+  ✔ replaces inherited child boundaries with the fanout boundary when authorized (0.321125ms)
+  ✔ replaces inherited fanout boundaries with the strict boundary when fanout is not authorized (0.227375ms)
+  ✔ keeps explicitly injected skill content when inherited skills are stripped (0.172584ms)
+  ✔ keeps configured lazy skill references when inherited skills are stripped (0.130833ms)
+  ✔ strips the subagent orchestration skill even when inherited skills remain (4.150417ms)
+  ✔ strips explicit dm-subagents skill injection from child prompts (0.065709ms)
+  ✔ strips parent-only subagent custom messages from forked child context (0.099958ms)
+  ✔ strips prior parent subagent tool calls and results from forked child context (0.07525ms)
+  ✔ preserves live nested subagent calls and results in fanout child context (0.057584ms)
+  ✔ defers native supervisor registration until runtime events and respects installed pi-intercom tools (0.232208ms)
+  ✔ keeps installed pi-intercom while filling only a missing child contact_supervisor tool (0.084583ms)
+  ✔ registers native supervisor tools at runtime when pi-intercom is absent (0.078292ms)
+  ✔ sets the child intercom session name from env during agent startup (0.059334ms)
+  ✔ rewrites the final child-visible prompt through before_agent_start (0.082167ms)
+  ✔ uses the fanout boundary through before_agent_start when fanout env is set (0.0585ms)
+  ✔ filters parent-only artifacts from polluted fork context while preserving ordinary history (0.077625ms)
+  ✔ does not rewrite child context when no parent-only artifacts are present (0.042375ms)
+✔ subagent prompt runtime (16.456ms)
+▶ resolveTempScopeId
+  ✔ prefers uid when available (0.266667ms)
+  ✔ falls back to environment usernames when uid is unavailable (0.110042ms)
+  ✔ falls back to os.userInfo when environment is missing (0.052875ms)
+  ✔ falls back to home path when os.userInfo throws (0.922583ms)
+✔ resolveTempScopeId (1.790834ms)
+▶ shared temp paths
+  ✔ anchors shared temp directories under one scoped root (0.284959ms)
+  ✔ writes async config files under the same scoped temp root (0.142416ms)
+✔ shared temp paths (0.555042ms)
+▶ tool-budget module
+  ✔ defaults block tools to read/search tools (0.553458ms)
+  ✔ accepts soft and wildcard block (0.060333ms)
+  ✔ rejects unsafe configs (0.071166ms)
+  ✔ serializes and decodes env config (0.091291ms)
+  ✔ tracks state and block decisions (0.090417ms)
+  ✔ formats user-facing budget messages (0.09125ms)
+✔ tool-budget module (1.371917ms)
+(node:44651) ExperimentalWarning: Transform Types is an experimental feature and might change at any time
+(Use `node --trace-warnings ...` to show where the warning was created)
+(node:47235) ExperimentalWarning: Transform Types is an experimental feature and might change at any time
+(Use `node --trace-warnings ...` to show where the warning was created)
+(node:49839) ExperimentalWarning: Transform Types is an experimental feature and might change at any time
+(Use `node --trace-warnings ...` to show where the warning was created)
+[dm-subagents] subagent-tool-description.md was not found or valid for toolDescriptionMode "custom"; using full description.
+(node:50811) ExperimentalWarning: Transform Types is an experimental feature and might change at any time
+(Use `node --trace-warnings ...` to show where the warning was created)
+[dm-subagents] Ignoring invalid toolDescriptionMode "tiny"; expected "full", "compact", or "custom".
+(node:51753) ExperimentalWarning: Transform Types is an experimental feature and might change at any time
+(Use `node --trace-warnings ...` to show where the warning was created)
+▶ registered subagent tool description
+  ✔ keeps full mode safe and free of hardcoded builtin agent names (1.2405ms)
+  ✔ offers a compact mode that keeps safety-critical guidance (0.4515ms)
+  ✔ renders a custom project description with placeholders and mandatory safety guidance (2.286084ms)
+  ✔ appends full safety guidance when custom prose only includes the safety heading (0.5865ms)
+  ✔ keeps mandatory safety guidance last when custom prose embeds it before an override (0.469125ms)
+  ✔ falls back to full mode when custom mode has no valid file (0.270541ms)
+  ✔ falls back to full mode when toolDescriptionMode is invalid (0.0705ms)
+  ✔ registers full, compact, custom, and fallback descriptions from extension config (4577.001333ms)
+✔ registered subagent tool description (4583.355375ms)
+▶ sumResultsUsage
+  ✔ aggregates full child usage (0.471875ms)
+✔ sumResultsUsage (0.789292ms)
+▶ sumResultsCost
+  ✔ aggregates input tokens, output tokens, and cost (0.1065ms)
+  ✔ returns zero totals when all aggregated fields are zero (0.04675ms)
+  ✔ includes attached nested subagent costs (0.053792ms)
+✔ sumResultsCost (0.283083ms)
+▶ ts-loader
+  ✔ rewrites .js imports when the parent path contains spaces (3.178667ms)
+✔ ts-loader (3.747208ms)
+▶ turn-budget module
+  ▶ DEFAULT_TURN_BUDGET_GRACE_TURNS
+    ✔ defaults to one grace turn (0.239083ms)
+  ✔ DEFAULT_TURN_BUDGET_GRACE_TURNS (0.486792ms)
+  ▶ appendTurnBudgetSystemPrompt
+    ✔ returns the system prompt unchanged when no budget is configured (2.010291ms)
+    ✔ appends a turn-budget block after an existing system prompt (0.17475ms)
+    ✔ emits only the turn-budget block when the system prompt is empty (0.055959ms)
+    ✔ trims surrounding whitespace from the existing system prompt before joining (0.052166ms)
+    ✔ pluralizes maxTurns for single and multiple turns (0.050209ms)
+    ✔ pluralizes graceTurns for one versus many or zero (0.053916ms)
+  ✔ appendTurnBudgetSystemPrompt (2.53375ms)
+  ▶ turnBudgetSoftNote
+    ✔ describes the wrap-up request with soft limit and grace values (0.072625ms)
+    ✔ pluralizes the turn count (0.085667ms)
+  ✔ turnBudgetSoftNote (0.226375ms)
+  ▶ turnBudgetExceededMessage
+    ✔ describes the hard abort with soft limit plus grace (0.077084ms)
+    ✔ pluralizes the turn count (0.0615ms)
+  ✔ turnBudgetExceededMessage (0.18475ms)
+  ▶ formatTurnBudgetOutput
+    ✔ prefixes the abort message onto captured partial output (0.046417ms)
+    ✔ preserves multi-line partial output verbatim (0.031875ms)
+    ✔ returns only the message when there is no partial output (0.024833ms)
+  ✔ formatTurnBudgetOutput (0.148792ms)
+  ▶ initialTurnBudgetState
+    ✔ starts within budget at turn zero and copies the resolved budget (0.347041ms)
+  ✔ initialTurnBudgetState (0.371791ms)
+  ▶ turnBudgetState
+    ✔ marks a wrap-up request without recording an exceeded turn (0.257167ms)
+    ✔ marks an exceeded budget and records the turn where it happened (0.176042ms)
+    ✔ records the wrap-up request at the soft-limit turn even when exceeded (0.035875ms)
+  ✔ turnBudgetState (0.534584ms)
+  ▶ shouldAbortForTurnBudget
+    ✔ allows a terminal assistant response on the final grace turn (0.048458ms)
+    ✔ aborts a non-terminal assistant response on the final grace turn (0.022875ms)
+    ✔ aborts turns beyond the grace window even if the message is terminal (0.01975ms)
+    ✔ allows a terminal response at the soft limit when grace turns are zero (0.021708ms)
+  ✔ shouldAbortForTurnBudget (0.151042ms)
+  ▶ turn budget state transitions
+    ✔ walks within-budget -> wrap-up-requested -> exceeded as turns accrue (0.041208ms)
+    ✔ can represent immediate exceedance when grace turns are zero (0.169958ms)
+  ✔ turn budget state transitions (0.240792ms)
+✔ turn-budget module (5.210542ms)
+▶ wrapForkTask
+  ✔ wraps task with default preamble (0.368209ms)
+  ✔ returns task unchanged when disabled (0.059583ms)
+  ✔ is idempotent for already wrapped tasks (0.044208ms)
+✔ wrapForkTask (0.826792ms)
+▶ wait tool
+  ✔ returns immediately when there is nothing to wait for (1.489959ms)
+  ✔ with all:true, resolves once every active run reaches a terminal state (31.29225ms)
+  ✔ wakes when a run needs attention, not only on completion (29.333709ms)
+  ✔ reports runs that already need attention before waiting starts (20.010958ms)
+  ✔ by default returns as soon as the FIRST run finishes, leaving the rest in flight (27.525958ms)
+  ✔ only waits for runs belonging to the current session (0.650291ms)
+  ✔ summarizes only runs that were active when waiting began (13.995458ms)
+  ✔ can target a single run by id prefix (26.885125ms)
+  ✔ rejects ambiguous id prefixes but lets exact ids win (28.806459ms)
+  ✔ times out while runs are still active and reports them (19.029709ms)
+  ✔ resolves early when the turn is aborted (19.1645ms)
+  ✔ wakes immediately on an event bus emission instead of waiting the poll interval (29.25025ms)
+  ✔ still resolves via poll when no event bus is provided (fallback) (14.073292ms)
+✔ wait tool (262.520084ms)
+▶ nested widget rendering
+  ✔ uses aggregate lines when collapsed and full child rows when expanded (8.902125ms)
+  ✔ collapses descendants beyond the nested depth budget (0.232166ms)
+  ✔ shows running descendants even after the parent step is complete (4.266417ms)
+  ✔ degrades stale child summaries to id and state (0.179417ms)
+  ✔ rerenders when only nested state changes (0.122583ms)
+✔ nested widget rendering (14.122208ms)
+▶ nested child DM process visibility
+  ✔ hides foreground child DM process windows on Windows (0.408417ms)
+  ✔ hides background child DM process windows on Windows (0.201209ms)
+✔ nested child DM process visibility (0.988209ms)
+▶ workflow graph snapshots
+  ✔ maps sequential chains with phase, label, output name, and stable flat indexes (0.684583ms)
+  ✔ maps a long alternating worker and reviewer chain without losing indexes (0.130709ms)
+  ✔ maps parallel chain steps with stable group and child indexes (0.104125ms)
+  ✔ marks partially completed parallel groups as running (0.066583ms)
+  ✔ summarizes mixed terminal parallel group statuses with explicit precedence (0.058708ms)
+  ✔ uses dynamic group status overrides for empty or aggregate-failure fanout states (0.206959ms)
+✔ workflow graph snapshots (1.660417ms)
+▶ worktree
+  ✔ createWorktrees returns expected structure (265.730792ms)
+  ✔ createWorktrees maps subdirectory cwd to each agentCwd (176.059ms)
+  ✔ previews expected worktree agent cwd for repository subdirectories (107.758458ms)
+  ✔ creates worktrees under a configured base directory (145.799625ms)
+  ✔ uses DM_SUBAGENTS_WORKTREE_DIR when no base directory is configured (152.853584ms)
+  ✔ createWorktrees rejects dirty repositories (96.845292ms)
+  ✔ findWorktreeTaskCwdConflict allows omitted or matching task cwd values (0.192833ms)
+  ✔ findWorktreeTaskCwdConflict treats relative task cwd values as relative to the shared cwd (0.085542ms)
+  ✔ findWorktreeTaskCwdConflict returns the first conflicting task cwd (0.175709ms)
+  ✔ diffWorktrees captures committed, modified, and new files without staging the node_modules symlink (204.110208ms)
+  ✔ cleanupWorktrees removes worktrees and branches (211.99575ms)
+  ✔ createWorktrees creates node_modules symlink when node_modules exists (144.5175ms)
+  ✔ diffWorktrees preserves a tracked node_modules symlink (207.598708ms)
+  ✔ runs a repo-relative worktree setup hook and records synthetic paths (440.063917ms)
+  ✔ runs an absolute worktree setup hook path (412.845667ms)
+  ✔ rejects bare command names for worktree setup hooks (99.324917ms)
+  ✔ rejects tracked synthetic paths from hook output (413.532291ms)
+  ✔ rejects absolute synthetic paths from hook output (427.839334ms)
+  ✔ excludes hook-created synthetic files from captured patch output (454.817875ms)
+  ✔ cleans up created worktrees when a later hook setup fails (482.614125ms)
+  ✔ fails when the hook exceeds the configured timeout (183.726709ms)
+✔ worktree (4629.327542ms)
+ℹ tests 974
+ℹ suites 169
+ℹ pass 974
+ℹ fail 0
+ℹ cancelled 0
+ℹ skipped 0
+ℹ todo 0
+ℹ duration_ms 8176.840042