git:20261012.b0585e7 to git:20261024.08e301f
67 added, 10 removed. Audit A to A.
- # Atlas
+ ---
+ name: atlas-agent
+ description: Developer reference for the Atlas todo-list orchestrator agent -- model variants, prompt sections, and routing.
+ ---
- `agent.ts` creates the primary orchestration agent with its existing settings.
- Its model core is generated by the public Senpi builders through
- `../native-model-prompt.ts`. Existing Atlas role markdown and dynamic injections
- retain the checkbox protocol, available agents, categories and skills.
- Role variants in prompts-core are not copies of Native model cores.
+ # src/agents/atlas/ -- Todo-List Orchestrator
- `prompt-section-builder.ts` supplies shared category descriptions.
- The runtime-injection test protects live context data; the parent catalog
- matrix proves rendered Native content parity. Per-request frame replacement
- handles runtime model switches while preserving the role and user appends.
+ **Generated:** 2026-05-18
+
+ ## OVERVIEW
+
+ 5 TypeScript files plus 8 markdown prompt variants in `packages/prompts-core/prompts/atlas/`. Atlas agent -- todo-list orchestrator that delegates via `task()` to complete every checkbox in a plan until fully done. Mode `primary`. Color `#10B981`.
+
+ ## FILES
+
+ | File | Purpose |
+ |------|---------|
+ | `agent.ts` | `createAtlasAgent()` factory, prompts-core variant loading, runtime placeholder injection, `OrchestratorContext` |
+ | `index.ts` | Barrel exports |
+ | `prompt-section-builder.ts` | Composes category, agent, skills, and decision matrix sections |
+ | `prompt-runtime-injection.test.ts` | Runtime placeholder-resolution regression tests |
+ | `prompt-routing.test.ts` | Model-variant routing tests |
+ | `packages/prompts-core/prompts/atlas/default.md` | Default/Claude markdown prompt variant |
+ | `packages/prompts-core/prompts/atlas/gpt.md` | GPT-optimized markdown prompt variant |
+ | `packages/prompts-core/prompts/atlas/gemini.md` | Gemini-optimized markdown prompt variant |
+ | `packages/prompts-core/prompts/atlas/kimi.md` | Kimi K2.x markdown prompt variant |
+ | `packages/prompts-core/prompts/atlas/kimi-k2-7.md` | Kimi K2.7 markdown prompt variant |
+ | `packages/prompts-core/prompts/atlas/kimi-k3.md` | Kimi K3-native markdown prompt variant |
+ | `packages/prompts-core/prompts/atlas/opus-4-7.md` | Claude Opus 4.7 markdown prompt variant |
+ | `packages/prompts-core/prompts/atlas/glm.md` | GLM-optimized markdown prompt variant |
+
+ ## MODEL VARIANT ROUTING
+
+ Parent `agent.ts` calls `resolveVariant()` from `@oh-my-opencode/prompts-core` against `atlasPromptVariants`:
+ - GPT family -> `gpt.md`
+ - Gemini family -> `gemini.md`
+ - Kimi K3.1 -> `kimi-k3-1.md` (its own copy of `kimi-k3.md`, checked before K3)
+ - Kimi K3 -> `kimi-k3.md` (checked first among the rest of Kimi)
+ - Kimi K2.7 -> `kimi-k2-7.md`
+ - Kimi K2.x family -> `kimi.md`
+ - GLM family -> `glm.md`
+ - Claude Opus 4.7 -> `opus-4-7.md`
+ - Default -> `default.md` (Claude Sonnet 5 family)
+
+ `atlasPromptVariants` is ordered with `opus-4-7` before `default` so the specific Claude Opus 4.7 route wins before the generic fallback.
+
+ ## RUNTIME INJECTION
+
+ The markdown files keep live OpenCode sections as placeholders. `agent.ts` resolves them through `loadPrompt()` runtime injections:
+ - `{CATEGORY_SECTION}` -> `buildCategorySection()`
+ - `{AGENT_SECTION}` -> `buildAgentSelectionSection()`
+ - `{DECISION_MATRIX}` -> `buildDecisionMatrix()`
+ - `{SKILLS_SECTION}` -> `buildSkillsSection()`
+ - `{{CATEGORY_SKILLS_DELEGATION_GUIDE}}` -> `buildCategorySkillsDelegationGuide()`
+
+ `prompt-section-builder.ts` remains the resolver implementation in `src/` because it depends on live category, agent, and skill state.
+
+ ## KEY BEHAVIORS
+
+ - Mode: `primary` (respects UI model selection)
+ - Temperature: 0.1
+ - Default model: `claude-sonnet-5`
+ - Denied tools: `task`, `call_omo_agent` (Atlas delegates; it does not run subagents directly)
+ - Checkbox enforcement in prompts
+ - Auto-continue: never asks user for approval between plan steps
+ - Parallel fan-out by default; sequential only for named blocking dependencies
+ - Post-delegation rule: edit plan checkbox, read plan to confirm, then dispatch next task
+ - Registered via `createAtlasAgent` in `src/agents/builtin-agents/atlas-agent.ts`
+ - Markdown prompts are imported with Bun's `.md` text loader so Atlas prompt content is bundled into `dist/index.js`.