feat(core): append model identity to system prompts via plugin

packages/core/src/plugin/identity.ts

@@ -0,0 +1,40 @@
+export * as IdentityPlugin from "./identity.js"
+
+import { SystemPart } from "@opencode/ai"
+import { define } from "@opencode/plugin/effect/plugin"
+import type { SessionHooks } from "@opencode/plugin/effect/session"
+import { Model } from "@opencode/schema/model"
+import { Effect } from "effect"
+
+export function identity(model: { readonly provider: string; readonly name: string; readonly ref: Model.Ref }) {
+  return [
+    "# Your Model",
+    `- Provider: ${model.provider}`,
+    `- Name: ${model.name}`,
+    `- ID: ${model.ref.providerID}/${model.ref.id}`,
+  ].join("\n")
+}
+
+export const Plugin = define({
+  id: "opencode.prompt.identity",
+  effect: Effect.fn("IdentityPlugin")(function* (ctx) {
+    const hook = (event: SessionHooks["context"]) =>
+      Effect.gen(function* () {
+        const model =
+          (yield* ctx.model.list()).data.find(
+            (model) => model.providerID === event.model.providerID && model.id === event.model.id,
+          ) ?? Model.Info.default(event.model.providerID, event.model.id)
+        const provider = (yield* ctx.provider.list()).data.find((provider) => provider.id === event.model.providerID)
+        event.system.splice(
+          1,
+          0,
+          SystemPart.make(
+            identity({ provider: provider?.name ?? event.model.providerID, name: model.name, ref: event.model }),
+          ),
+        )
+      }).pipe(Effect.catch(() => Effect.void))
+    yield* ctx.session.hook("context", hook)
+    yield* ctx.session.hook("compaction", hook)
+    yield* ctx.session.hook("generate", hook)
+  }),
+})

packages/core/src/plugin/internal.ts

@@ -84,6 +84,7 @@ import { WriteTool } from "../tool/plugin/write.js"
 import { AgentPlugin } from "./agent.js"
 import BrowserPlugin from "@opencode/plugin-browser"
 import { CommandPlugin } from "./command.js"
+import { IdentityPlugin } from "./identity.js"
 import { PlanPlugin } from "./plan.js"
 import { ModelsDevPlugin } from "./models-dev.js"
 import { McpCodeModeExclusionPlugin } from "./mcp-codemode-exclusion.js"
@@ -216,6 +217,7 @@ const pre = [
   PatchTool.Plugin,
   // Render model prompts after the patch plugin selects the available editing tools.
   ...OptimizePlugin.Plugins,
+  IdentityPlugin.Plugin,
   EditTool.Plugin,
   GlobTool.Plugin,
   GrepTool.Plugin,

packages/core/src/plugin/system-prompt/gpt-astra.txt

@@ -1,4 +1,4 @@
-You are an AI agent powered by OpenCode, a coding agent harness. Help the user accomplish their goals using the tools you have available.
+You are an AI agent running in OpenCode, a coding agent harness. Help the user accomplish their goals using the tools you have available.
 
 # Harness
 - Responses are rendered as GitHub-flavored Markdown.

packages/core/src/plugin/system-prompt/gpt.txt

@@ -1,4 +1,4 @@
-You are an AI agent powered by OpenCode, a coding agent harness. Help the user accomplish their goals using the tools you have available.
+You are an AI agent running in OpenCode, a coding agent harness. Help the user accomplish their goals using the tools you have available.
 
 # Harness
 - Responses are rendered as GitHub-flavored Markdown.

packages/core/src/session/runner/prompt/system.txt

@@ -1,4 +1,4 @@
-You are an AI agent powered by OpenCode, a coding agent harness. Help the user accomplish their goals using the tools you have available.
+You are an AI agent running in OpenCode, a coding agent harness. Help the user accomplish their goals using the tools you have available.
 
 # Harness
 - Responses are rendered as GitHub-flavored Markdown.

packages/core/test/fixtures/recordings/session-runner/openai-chat-streams-text.json

@@ -13,7 +13,7 @@
         "headers": {
           "content-type": "application/json"
         },
-        "body": "{\"model\":\"gpt-4o-mini\",\"messages\":[{\"role\":\"system\",\"content\":\"You are an AI agent powered by OpenCode, a coding agent harness. Help the user accomplish their goals using the tools you have available.\\n\\n# Harness\\n- Responses are rendered as GitHub-flavored Markdown.\\n- `<system-reminder>` blocks are harness instructions, not user-authored content. Read and follow them.\\n- Prefer parallelizing independent tool calls.\\n\\n\\n# Communication\\n\\nUse clear file paths when referring to files. Keep responses clear and concise, and avoid unnecessary technical jargon.\\n\\n## Intermediate Commentary\\n\\nAs you work, you send messages to the commentary channel. These are how you collaborate with the user while you work: stating assumptions and providing updates. Keep them concise and quickly scannable, and send them only when they add real information, such as a discovery, a tradeoff, or a blocker. Do not narrate routine reads, searches, or edits.\\n\\nBy default, treat new messages received during ongoing work as steering the active task rather than replacing it. Incorporate corrections and constraints, and answer questions briefly in commentary before continuing. Replace the task only when the user clearly cancels it or requests an incompatible objective.\\n\\nDo not put a final response, such as a blocking or clarifying question, in the commentary channel. The final answer must always be fully self-contained.\\n\\n## Final Answer\\n\\nIn the final answer, lead with the outcome, not the steps you took to reach it. Cover the most important information, use only as much structure as the answer needs, and avoid long-winded explanations unless necessary. Include technical detail only where it helps.\\n\\n# Working in codebases\\n- Keep changes consistent with the structure, naming, style, and patterns of the surrounding code.\\n- Treat unfamiliar files or changes as potential user work and investigate before deleting or overwriting them.\\n\\n# Delegation\\n\\nDo not spawn subagents unless the user or applicable AGENTS.md/skill instructions explicitly ask for subagents, delegation, or parallel agent work.\\n\\n# Destructive actions\\n\\nDo not revert, reset, or discard changes you did not make. Never run destructive commands such as `git reset --hard`, `git checkout --`, or recursive deletes on broad paths unless the user clearly asked for that operation; if the target or scope is unclear, ask first. Prefer non-interactive git commands.\\n\\n# Autonomy\\n\\nDo not infer authorization for work beyond the user's request. Assumptions that help you make progress are fine as long as they stay within the user's intent and the scope of the task.\\n\"},{\"role\":\"user\",\"content\":\"Say hello in one short sentence.\"}],\"stream\":true,\"stream_options\":{\"include_usage\":true},\"store\":false,\"prompt_cache_key\":\"ses_runner_recorded\",\"max_completion_tokens\":20,\"temperature\":0}"
+        "body": "{\"model\":\"gpt-4o-mini\",\"messages\":[{\"role\":\"system\",\"content\":\"You are an AI agent running in OpenCode, a coding agent harness. Help the user accomplish their goals using the tools you have available.\\n\\n# Harness\\n- Responses are rendered as GitHub-flavored Markdown.\\n- `<system-reminder>` blocks are harness instructions, not user-authored content. Read and follow them.\\n- Prefer parallelizing independent tool calls.\\n\\n\\n# Communication\\n\\nUse clear file paths when referring to files. Keep responses clear and concise, and avoid unnecessary technical jargon.\\n\\n## Intermediate Commentary\\n\\nAs you work, you send messages to the commentary channel. These are how you collaborate with the user while you work: stating assumptions and providing updates. Keep them concise and quickly scannable, and send them only when they add real information, such as a discovery, a tradeoff, or a blocker. Do not narrate routine reads, searches, or edits.\\n\\nBy default, treat new messages received during ongoing work as steering the active task rather than replacing it. Incorporate corrections and constraints, and answer questions briefly in commentary before continuing. Replace the task only when the user clearly cancels it or requests an incompatible objective.\\n\\nDo not put a final response, such as a blocking or clarifying question, in the commentary channel. The final answer must always be fully self-contained.\\n\\n## Final Answer\\n\\nIn the final answer, lead with the outcome, not the steps you took to reach it. Cover the most important information, use only as much structure as the answer needs, and avoid long-winded explanations unless necessary. Include technical detail only where it helps.\\n\\n# Working in codebases\\n- Keep changes consistent with the structure, naming, style, and patterns of the surrounding code.\\n- Treat unfamiliar files or changes as potential user work and investigate before deleting or overwriting them.\\n\\n# Delegation\\n\\nDo not spawn subagents unless the user or applicable AGENTS.md/skill instructions explicitly ask for subagents, delegation, or parallel agent work.\\n\\n# Destructive actions\\n\\nDo not revert, reset, or discard changes you did not make. Never run destructive commands such as `git reset --hard`, `git checkout --`, or recursive deletes on broad paths unless the user clearly asked for that operation; if the target or scope is unclear, ask first. Prefer non-interactive git commands.\\n\\n# Autonomy\\n\\nDo not infer authorization for work beyond the user's request. Assumptions that help you make progress are fine as long as they stay within the user's intent and the scope of the task.\\n\\n# Your Model\\n- Provider: openai\\n- Name: gpt-4o-mini\\n- ID: openai/gpt-4o-mini\"},{\"role\":\"user\",\"content\":\"Say hello in one short sentence.\"}],\"stream\":true,\"stream_options\":{\"include_usage\":true},\"store\":false,\"prompt_cache_key\":\"ses_runner_recorded\",\"max_completion_tokens\":20,\"temperature\":0}"
       },
       "response": {
         "status": 200,

packages/core/test/plugin/host.ts

@@ -252,6 +252,14 @@ export function providerHost(providers: Provider.Interface): Plugin.Context["pro
   }
 }
 
+/** Empty catalog for hosts that only need provider lookups to fall back. */
+export const noProviders: Plugin.Context["provider"] = {
+  list: () => Effect.succeed(located([])),
+  get: () => Effect.die("unused provider.get"),
+  transform: () => Effect.die("unused provider.transform"),
+  reload: () => Effect.die("unused provider.reload"),
+}
+
 export function modelHost(models: Model.Interface): Plugin.Context["model"] {
   return {
     list: () => models.available().pipe(Effect.map(located)),

packages/core/test/plugin/identity.test.ts

@@ -0,0 +1,73 @@
+import { expect, test } from "bun:test"
+import { SystemPart } from "@opencode/ai"
+import { Agent } from "@opencode/core/agent"
+import { IdentityPlugin } from "@opencode/core/plugin/identity"
+import { Plugin } from "@opencode/core/plugin"
+import { PluginHooks } from "@opencode/core/plugin/hooks"
+import { PluginHost } from "@opencode/core/plugin/host"
+import { Session } from "@opencode/core/session"
+import type { SessionHooks } from "@opencode/plugin/effect/session"
+import { Model } from "@opencode/schema/model"
+import { Provider } from "@opencode/core/provider"
+import { Effect } from "effect"
+import { testEffect } from "../lib/effect"
+import { PluginTestLayer } from "./fixture"
+
+const it = testEffect(PluginTestLayer)
+
+test("formats the model identity part", () => {
+  expect(
+    IdentityPlugin.identity({
+      provider: "OpenAI",
+      name: "GPT-4o mini",
+      ref: Model.Ref.make({ providerID: Provider.ID.make("openai"), id: Model.ID.make("gpt-4o-mini") }),
+    }),
+  ).toBe(["# Your Model", "- Provider: OpenAI", "- Name: GPT-4o mini", "- ID: openai/gpt-4o-mini"].join("\n"))
+})
+
+const identity = (provider: string, name: string, id: string) =>
+  ["# Your Model", `- Provider: ${provider}`, `- Name: ${name}`, `- ID: test/${id}`].join("\n")
+
+const context = (id: string): SessionHooks["context"] => ({
+  sessionID: Session.ID.make("ses_model_identity"),
+  agent: Agent.ID.make("build"),
+  model: Model.Ref.make({ providerID: Provider.ID.make("test"), id: Model.ID.make(id) }),
+  system: [SystemPart.make("Agent prompt"), SystemPart.make("Initial context")],
+  messages: [],
+  tools: {},
+  options: {},
+})
+
+it.effect("inserts the structured model block after the agent prompt", () =>
+  Effect.gen(function* () {
+    const catalog = yield* Provider.Service
+    const hooks = yield* PluginHooks.Service
+    const plugins = yield* Plugin.Service
+    const pluginHost = yield* PluginHost.make(plugins)
+    yield* catalog.transform((editor) => {
+      editor.update(Provider.ID.make("test"), (provider) => {
+        provider.name = "Test Provider"
+      })
+      editor.models.update(Provider.ID.make("test"), Model.ID.make("meta/muse-spark-1.1"), (model) => {
+        model.name = "Muse Spark"
+      })
+    })
+    yield* IdentityPlugin.Plugin.effect(pluginHost)
+
+    const named = context("meta/muse-spark-1.1")
+    yield* hooks.trigger("session", "context", named)
+    expect(named.system.map((part) => part.text)).toEqual([
+      "Agent prompt",
+      identity("Test Provider", "Muse Spark", "meta/muse-spark-1.1"),
+      "Initial context",
+    ])
+
+    const fallback = context("unknown-model")
+    yield* hooks.trigger("session", "context", fallback)
+    expect(fallback.system.map((part) => part.text)).toEqual([
+      "Agent prompt",
+      identity("Test Provider", "unknown-model", "unknown-model"),
+      "Initial context",
+    ])
+  }),
+)

packages/core/test/session-runner-recorded.test.ts

@@ -36,6 +36,7 @@ import { PluginSupervisor } from "@opencode/core/plugin/supervisor"
 import { Plugin } from "@opencode/core/plugin"
 import { PluginHooks } from "@opencode/core/plugin/hooks"
 import { OptimizePlugin } from "@opencode/core/plugin/optimize"
+import { IdentityPlugin } from "@opencode/core/plugin/identity"
 import { describe, expect } from "bun:test"
 import { eq } from "drizzle-orm"
 import { Effect, Layer } from "effect"
@@ -44,7 +45,7 @@ import { testEffect } from "./lib/effect"
 import { LocationServiceMap } from "@opencode/core/location-service-map"
 import { promptLocationNode } from "./fixture/prompt-location"
 import { permissionLayer } from "./lib/permission"
-import { agentHost, modelHost, host } from "./plugin/host"
+import { agentHost, modelHost, host, noProviders } from "./plugin/host"
 
 const cassetteName = "session-runner/openai-chat-streams-text"
 const cassetteDirectory = path.resolve(import.meta.dir, "fixtures/recordings")
@@ -184,9 +185,11 @@ describe("SessionRunnerLLM recorded", () => {
       const pluginHost = host({
         agent: agentHost(agents),
         model: modelHost(models),
+        provider: noProviders,
         session: { hook: (name, callback) => hooks.register("session", name, callback) },
       })
       yield* Effect.forEach(OptimizePlugin.Plugins, (plugin) => plugin.effect(pluginHost), { discard: true })
+      yield* IdentityPlugin.Plugin.effect(pluginHost)
       const { db } = yield* Database.Service
       yield* db
         .insert(ProjectTable)