diff --git a/packages/opencode/src/altimate/telemetry/error-detail.ts b/packages/opencode/src/altimate/telemetry/error-detail.ts
index 3d2ef5e78d..781c2855aa 100644
--- a/packages/opencode/src/altimate/telemetry/error-detail.ts
+++ b/packages/opencode/src/altimate/telemetry/error-detail.ts
@@ -392,6 +392,18 @@ export function templateErrorMessage(raw: string): string {
   return out.join(" ").slice(0, MAX_MESSAGE)
 }
 
+/**
+ * True only for the exact messages of PermissionV1.RejectedError / CorrectedError (the user
+ * declining a permission prompt, with or without feedback), optionally prefixed by the
+ * error name from toString(). That is a user decision, not a tool fault; config-rule
+ * denials and real "permission denied" failures do not match and are still reported.
+ */
+export function isUserPermissionRejection(raw: string): boolean {
+  return /^(?:[A-Za-z.]*Error: )?The user rejected permission to use this specific tool call(?:\.$| with the following feedback: )/.test(
+    raw,
+  )
+}
+
 /** Hash input is unchanged from before the detail fields existed, so fingerprints stay stable. */
 export function describeToolError(raw: string) {
   return {
diff --git a/packages/opencode/src/session/prompt.ts b/packages/opencode/src/session/prompt.ts
index 11a455551d..9fb67537d5 100644
--- a/packages/opencode/src/session/prompt.ts
+++ b/packages/opencode/src/session/prompt.ts
@@ -2289,6 +2289,17 @@ export namespace SessionPrompt {
         // altimate_change end
         if (toolChain.length < 50) toolChain.push(part.tool)
         const isError = part.state?.status === "error"
+        // altimate_change start — a user declining a permission prompt is a decision, not a tool fault: no error fingerprint
+        if (
+          isError &&
+          part.state.status === "error" &&
+          typeof part.state.error === "string" &&
+          ErrorDetail.isUserPermissionRejection(part.state.error)
+        ) {
+          toolErrorCount++
+          continue
+        }
+        // altimate_change end
         if (isError) {
           toolErrorCount++
           // Flush previous unrecovered error before recording new one
diff --git a/packages/opencode/src/tool/tool.ts b/packages/opencode/src/tool/tool.ts
index 4558050b1d..0c9cd5d982 100644
--- a/packages/opencode/src/tool/tool.ts
+++ b/packages/opencode/src/tool/tool.ts
@@ -8,6 +8,7 @@ import { Agent } from "@/agent/agent"
 // altimate_change start — telemetry instrumentation + legacy zod-tool adapter
 import z from "zod"
 import { Telemetry } from "../altimate/telemetry"
+import { ErrorDetail } from "../altimate/telemetry/error-detail"
 // altimate_change start — debug mode tool tracing
 import { traceToolCall } from "@/altimate/debug/mode"
 // altimate_change end
@@ -186,6 +187,8 @@ function altimateTrackError(id: string, args: unknown, ctx: Context, startTime:
       input_signature: Telemetry.computeInputSignature(args as Record<string, unknown>),
       error: maskedErrorMsg,
     })
+    // a user declining the permission prompt is a decision, not a core failure
+    if (ErrorDetail.isUserPermissionRejection(errorMsg)) return
     Telemetry.track({
       type: "core_failure",
       timestamp: Date.now(),
diff --git a/packages/opencode/test/altimate/permission-rejection-telemetry.test.ts b/packages/opencode/test/altimate/permission-rejection-telemetry.test.ts
new file mode 100644
index 0000000000..6c9c81b5c7
--- /dev/null
+++ b/packages/opencode/test/altimate/permission-rejection-telemetry.test.ts
@@ -0,0 +1,80 @@
+import { afterEach, describe, expect, spyOn, test } from "bun:test"
+import z from "zod"
+import { PermissionV1 } from "@opencode-ai/core/v1/permission"
+import { Tool } from "../../src/tool/tool"
+import { Telemetry } from "../../src/altimate/telemetry"
+import { ErrorDetail } from "../../src/altimate/telemetry/error-detail"
+import { PermissionNext } from "../../src/permission/next"
+import { initTool } from "./tool-fixture"
+
+const spies: Array<{ mockRestore(): void }> = []
+afterEach(() => {
+  for (const spy of spies.splice(0)) spy.mockRestore()
+})
+
+function ctx(ask: () => Promise<void>): any {
+  return {
+    sessionID: "ses_perm_reject",
+    messageID: "msg_perm_reject",
+    agent: "build",
+    abort: AbortSignal.any([]),
+    messages: [],
+    metadata() {},
+    ask,
+  }
+}
+
+async function run(error: Error) {
+  const events: Telemetry.Event[] = []
+  spies.push(spyOn(Telemetry, "track").mockImplementation((event) => void events.push(event)))
+  const tool = await initTool(
+    Tool.define("perm_fixture", {
+      description: "Permission rejection fixture",
+      parameters: z.object({}),
+      async execute(_args, c) {
+        await c.ask({ permission: "perm_fixture", patterns: ["*"], always: ["*"], metadata: {} })
+        return { title: "", output: "ok", metadata: {} }
+      },
+    }),
+  )
+  await expect(
+    tool.execute(
+      {},
+      ctx(async () => {
+        throw error
+      }),
+    ),
+  ).rejects.toBeDefined()
+  return events
+}
+
+describe("user permission rejection telemetry", () => {
+  test.each([
+    ["PermissionNext.RejectedError toString", new PermissionNext.RejectedError().toString()],
+    ["PermissionV1.RejectedError message", new PermissionV1.RejectedError().message],
+    ["PermissionV1.CorrectedError message", new PermissionV1.CorrectedError({ feedback: "use glob" }).message],
+    ["PermissionNext.CorrectedError toString", new PermissionNext.CorrectedError("no").toString()],
+  ])("%s is recognised as a user decision", (_label, raw) => {
+    expect(ErrorDetail.isUserPermissionRejection(raw)).toBe(true)
+  })
+
+  test.each([
+    new PermissionV1.DeniedError({ ruleset: [] }).message,
+    "EACCES: permission denied, open '/etc/shadow'",
+    "Access denied for user",
+    "tool failed: The user rejected permission to use this specific tool call.",
+  ])("other permission failures are still reported: %s", (raw) => {
+    expect(ErrorDetail.isUserPermissionRejection(raw)).toBe(false)
+  })
+
+  test("a rejected prompt emits no core_failure", async () => {
+    const events = await run(new PermissionNext.RejectedError())
+    expect(events.find((e) => e.type === "tool_call")).toMatchObject({ status: "error" })
+    expect(events.some((e) => e.type === "core_failure")).toBe(false)
+  })
+
+  test("a config-rule denial still emits core_failure", async () => {
+    const events = await run(new PermissionNext.DeniedError([]))
+    expect(events.some((e) => e.type === "core_failure")).toBe(true)
+  })
+})