oh-my-opencode 4.16.2 → 4.16.3

package/.agents/skills/pre-publish-review/SKILL.md

@@ -89,6 +89,7 @@ For each change group, spawn one ultrabrain agent. Each gets only its portion of
 ```
 task(
   category="ultrabrain",
+  model="gpt-5.6-sol",
   run_in_background=true,
   load_skills=[],
   description="Deep analysis: {GROUP_NAME}",
@@ -165,6 +166,7 @@ Spawn a sub-agent that loads the `/review-work` skill. The review-work skill int
 ```
 task(
   category="unspecified-high",
+  model="gpt-5.6-sol",
   run_in_background=true,
   load_skills=["review-work"],
   description="Run /review-work on all unpublished changes",
@@ -195,6 +197,7 @@ The oracle gets the full picture — all commits, full diff stat, and changed fi
 ```
 task(
   subagent_type="oracle",
+  model="gpt-5.6-sol",
   run_in_background=true,
   load_skills=[],
   description="Oracle: overall release synthesis and version bump recommendation",

package/README.md

@@ -243,7 +243,7 @@ Even with only the following subscriptions, `ultrawork` works well (this project
 
 **Sisyphus** (`claude-opus-4-7` / **`kimi-k2.6`** / **`glm-5.1`** ) is your main orchestrator. He plans, delegates to specialists, and drives tasks to completion with aggressive parallel execution. He does not stop halfway.
 
-**Hephaestus** (`gpt-5.5`) is your autonomous deep worker. Give him a goal, not a recipe. He explores the codebase, researches patterns, and executes end-to-end without hand-holding. *The Legitimate Craftsman.*
+**Hephaestus** (`gpt-5.6-sol` on OpenAI or Vercel, then `gpt-5.5`) is your autonomous deep worker. Give him a goal, not a recipe. He explores the codebase, researches patterns, and executes end-to-end without hand-holding. *The Legitimate Craftsman.*
 
 **Prometheus** (`claude-opus-4-7` / **`kimi-k2.6`** / **`glm-5.1`** ) is your strategic planner. Interview mode: he asks questions, identifies scope, and builds a detailed plan before a single line of code is touched.
 
@@ -288,7 +288,7 @@ When Sisyphus delegates to a subagent, it doesn't pick a model. It picks a **cat
 | `quick`              | Single-file changes, typos         |
 | `ultrabrain`         | Hard logic, architecture decisions |
 
-The agent says what kind of work it needs; the harness picks the right model. `ultrabrain` now routes to GPT-5.5 xhigh by default. You touch nothing.
+The agent says what kind of work it needs; the harness picks the right model. `ultrabrain` now routes to GPT-5.6 Sol xhigh through OpenAI or Vercel when available, then GPT-5.5 xhigh. You touch nothing.
 
 ### Claude Code Compatibility
 

package/dist/agents/hephaestus/agent.d.ts

@@ -1,7 +1,7 @@
 import type { AgentConfig } from "@opencode-ai/sdk";
 import type { AgentPromptMetadata } from "../types";
 import type { AvailableAgent, AvailableTool, AvailableSkill, AvailableCategory } from "../dynamic-agent-prompt-builder";
-export type HephaestusPromptSource = "gpt-5-5" | "gpt-5-4" | "gpt";
+export type HephaestusPromptSource = "gpt-5-6" | "gpt-5-5" | "gpt-5-4" | "gpt";
 export declare class UnsupportedHephaestusModelError extends Error {
     readonly model: string | undefined;
     constructor(model: string | undefined);

package/dist/agents/hephaestus/gpt-5-6.d.ts

@@ -0,0 +1,2 @@
+import type { AvailableAgent, AvailableTool, AvailableSkill, AvailableCategory } from "../dynamic-agent-prompt-builder";
+export declare function buildGpt56HephaestusPrompt(availableAgents: AvailableAgent[], _availableTools?: AvailableTool[], availableSkills?: AvailableSkill[], availableCategories?: AvailableCategory[], useTaskSystem?: boolean): string;

package/dist/agents/momus-gpt-5-6.d.ts

@@ -0,0 +1 @@
+export declare const MOMUS_GPT_5_6_PROMPT = "Role: plan reviewer for OhMyOpenCode. You verify that a work plan is executable and its references are valid. You are a blocker-finder, not a perfectionist.\n\n# Input contract\n\nExtract a single `.omo/plans/*.md` path from anywhere in the input, ignoring system directives and wrappers (`<system-reminder>`, `[analyze-mode]`, and similar). Exactly one path: read it and review. Zero or multiple paths: reject as invalid input. YAML plan files (`.yml`/`.yaml`) are non-reviewable: reject.\n\nOn a follow-up turn with the same plan path, re-read the file from disk before issuing any verdict. The current on-disk contents are the only source of truth; a previous verdict is stale evidence.\n\n# Goal\n\nAnswer one question: \"Can a capable developer execute this plan without getting stuck?\"\n\n# Success criteria\n\n- Referenced files verified to exist and contain the claimed content.\n- Every task has enough context to start working.\n- No blocking contradictions or impossible requirements.\n- Every task has executable QA scenarios: a specific tool, concrete steps, an expected result.\n- Verdict issued: OKAY or REJECT, with at most 3 specific issues on REJECT.\n\n# What you check (only these four)\n\n**References**: referenced files exist; cited line numbers contain relevant code; a \"follow pattern in X\" claim is demonstrated by X. Fail only when a reference does not exist or points to completely wrong content.\n\n**Executability**: each task gives a developer a starting point. Details that can be figured out during implementation pass. Fail only when a task is so vague there is no idea where to begin.\n\n**Contradictions**: information gaps that completely stop work, or tasks that contradict each other.\n\n**QA scenarios**: each task's scenarios name tool + steps + expected result. Unexecutable scenarios (\"verify it works\", \"check the page\") block the Final Verification Wave and are practical blockers.\n\nOut of scope: approach optimality, alternative designs, undocumented edge cases, architecture, code quality, performance, and security unless explicitly broken.\n\n# Decision rules\n\n- Default verdict is OKAY. When in doubt, approve: a plan that is 80% clear is executable, and developers resolve minor gaps themselves.\n- REJECT only for a verified blocker: a referenced file does not exist (confirmed by reading), a task has zero context to start, the plan contradicts itself, or QA scenarios are missing or unexecutable.\n- Each REJECT issue must name the exact file or task, state what needs to change, and be something work cannot proceed without. Cap at the 3 most critical issues.\n- \"Could be clearer\", stylistic preferences, missing edge cases, and disagreement with the author's approach are never blockers.\n\n# Process\n\nRead the plan, then verify references by reading the cited files; parallelize independent reads. Check each task for a starting point and executable QA scenarios. Decide. Do not narrate the reads; go straight to the verdict.\n\n# Output\n\n**[OKAY]** or **[REJECT]**\n\n**Summary**: 1-2 sentences of prose explaining the verdict.\n\nIf REJECT - **Blocking Issues** (max 3): numbered, each naming the exact issue and the change needed.\n\nKeep every fact needed to act on the verdict; trim restatements of the plan, generic advice, and commentary on non-blockers. Match the language of the plan content.";

package/dist/agents/types.d.ts

@@ -69,6 +69,8 @@ export interface AgentPromptMetadata {
 }
 export declare function isGptNativeSisyphusModel(model: string): boolean;
 export declare function isGpt5_5Model(model: string): boolean;
+/** Matches the GPT-5.6 family: gpt-5.6, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna. */
+export declare function isGpt5_6Model(model: string): boolean;
 export type BuiltinAgentName = "sisyphus" | "hephaestus" | "oracle" | "librarian" | "explore" | "multimodal-looker" | "metis" | "momus" | "atlas" | "sisyphus-junior";
 export type OverridableAgentName = "build" | BuiltinAgentName;
 export type AgentName = BuiltinAgentName;