Extract the tool description/category metadata into tool-description.ts and move argument normalization plus validation into tool-argument-preparation.ts. This keeps createDelegateTask focused on orchestration while preserving behavior and bringing tools.ts under the module LOC rule.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
Match opencode core 'task' tool behavior for metadata consistency:
1. Model fallback: When categoryModel/task.model/resumeModel is undefined,
fall back to parentContext.model so subagent metadata always includes
model info. Thread parentContext into executeSyncContinuation for parity.
2. Task ID consistency: unstable-agent-task was missing taskId and
backgroundTaskId in metadata. background_output used inconsistent
snake_case 'task_id' vs camelCase 'taskId' elsewhere. Standardize on
camelCase: taskId = sessionID (resume id), backgroundTaskId = bg task id.
Update text output blocks to use buildTaskMetadataBlock helper.
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)
Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
When weaker models (GLM-5, MiniMax) omit the description parameter on
delegate_task, the tool now auto-generates it from the first 4 words of
the prompt instead of throwing an error.
Changes:
- Schema: description is now optional (tool.schema.string().optional())
- Runtime: auto-generates from prompt when missing/empty/whitespace
- DelegateTaskArgs.description type stays as string (guaranteed by auto-gen)
- Tests: 3 new cases - missing/empty/explicit description handling
- Metadata title set after description resolution (correct ordering)
OpenCode's fromPlugin wrapper skips Zod validation for plugin tools, so
LLMs can omit required args like description without getting an error.
When Atlas orchestrates and the model omits description, it flows through
as undefined to manager.launch() and background task notifications show
'undefined' for all completed tasks.
Two fixes:
- Add runtime validation for description in delegate-task tool (matches
existing run_in_background and load_skills validation pattern)
- Defensive fallback in notification template: use task ID when
description is missing instead of rendering 'undefined'
Add support for object-style entries in fallback_models arrays, enabling
per-model configuration of variant, reasoningEffort, temperature, top_p,
maxTokens, and thinking settings.
- Zod schema for FallbackModelObject with full validation
- normalizeFallbackModels() and flattenToFallbackModelStrings() utilities
- Provider-agnostic model resolution pipeline with fallback chain
- Session prompt params state management
- Fallback chain construction with prefix-match lookup
- Integration across delegate-task, background-agent, and plugin layers
- #2741: Pass inheritedModel as fallback in subagent-resolver when user hasn't
configured an override, ensuring custom provider models take priority
- #2648: Fix getPlanProgress to treat plans with 0 checkboxes as incomplete
instead of complete (total > 0 && completed === total)
- #2779: Relax Hephaestus single-task guard to accept multi-step sub-tasks
from Atlas delegation, only rejecting genuinely independent tasks
Fixes#2741, fixes#2648, fixes#2779
Remove the auto-defaulting logic from PR #2420 that silently set
run_in_background=false when category/subagent_type/session_id was present.
The tool description falsely claimed 'Default: false' which misled agents
into omitting the parameter. Now the description says REQUIRED and the
validation always throws when the parameter is missing, with a clear
error message guiding the agent to retry with the correct value.
Reverts the behavioral change from #2420 while keeping the issue's
root cause (misleading description) fixed.
Problem: Agents frequently omit both 'category' and 'subagent_type' parameters
when calling the task() tool, causing validation failures. The JSON Schema
marks both as optional, and LLMs follow schema structure over description text.
Solution (Option A): Add aggressive visual warnings and failure-mode examples
to the tool description:
- ⚠️ CRITICAL warning header
- COMMON MISTAKE example showing what will FAIL
- CORRECT examples for both category and subagent_type usage
- Clear explanation that ONE must be provided
Tests: All 153 existing tests pass (no behavior change, only prompt improvement)
Category-based delegation (task(category='quick')) was broken because
SISYPHUS_JUNIOR_AGENT sent 'sisyphus-junior' to session.prompt but
config.agent keys are now display names ('Sisyphus-Junior').
- Use getAgentDisplayName() for SISYPHUS_JUNIOR_AGENT constant
- Replace hardcoded 'sisyphus-junior' strings in tools.ts with constant
- Update background-output local constants to use display names
- Add remapCommandAgentFields() to translate command agent fields
- Add raw-key fallback in tool-config-handler agentByKey()
LLMs sometimes pass load_skills as a serialized JSON string instead
of an array. Add defensive JSON.parse before validation to handle
this gracefully.
Fixes#1701
Community-reported-by: @omarmciver
Allow individual categories to be disabled via `disable: true` in
config. Introduce shared `mergeCategories()` utility to centralize
category merging and disabled filtering across all 7 consumption sites.
Category-based delegation should always route to sisyphus-junior even if
subagent_type is mistakenly provided, matching the original behavior and
preventing accidental bypass of category routing.
- Create available-models.ts for model availability checking
- Create model-selection.ts for category-to-model resolution logic
- Update category-resolver, subagent-resolver, and sync modules to import
from new focused modules instead of monolithic sources
- Add variant: max to ultrabrain's gemini-3-pro fallback entry
- Rename STRATEGIC_CATEGORY_PROMPT_APPEND to ULTRABRAIN_CATEGORY_PROMPT_APPEND
- Keep original strategic advisor prompt content (no micromanagement instructions)
- Update description: use only for genuinely hard tasks, give clear goals only
- Update tests to match renamed constant
PR #1227 incorrectly removed resolved.model from the userModel chain,
assuming it was bypassing the fallback chain. However, resolved.model
contained the category's DEFAULT_CATEGORIES model (e.g., quick ->
claude-haiku-4-5), not the main session model.
Without resolved.model, when connectedProvidersCache is null and
availableModels is empty, category model resolution falls through to
systemDefaultModel (opus) instead of using the category's default.
This fix restores the original priority:
1. User category model override
2. Category default model (from resolved.model)
3. sisyphusJuniorModel
4. Fallback chain
5. System default
When delegate_task uses subagent_type, extract the matched agent's model
object and pass it explicitly to session.prompt/manager.launch. This
ensures the model is always in the correct object format regardless of
how OpenCode handles string→object conversion for plugin-registered
agents.
Closes#1225
* fix(start-work): prevent overwriting session agent if already set; inherit parent model for subagent types
* fix(model): include variant in StoredMessage model structure for better context propagation
* fix(injector): include variant in model structure for hook message injection
- Remove resolved.model from userModel in tools.ts (was bypassing fallback chain)
- Use connected providers cache in model-resolver when availableModels is empty
- Allows proper provider selection (e.g., github-copilot instead of google)
- Block prometheus from delegating to itself via delegate_task
- Grant delegate_task permission to prometheus when called as subagent
- Other subagents still have delegate_task disabled
- Add timing.ts module for test-only timing configuration
- Replace hardcoded wait times with getTimingConfig()
- Enable all previously skipped tests (ralph-loop, session-state, delegate-task)
- Tests now complete in ~2s instead of timing out
- Add permission: [{ permission: 'question', action: 'deny' }] to session.create()
in background-agent and delegate-task for SDK-level blocking
- Add subagent-question-blocker hook as backup layer to intercept question tool
calls in tool.execute.before event
- Ensures subagents cannot ask questions to users and must work autonomously
YeonGyu-Kim