diff --git a/examples/acp-agent/cordis.snapshot-record.yml b/examples/acp-agent/cordis.snapshot-record.yml
new file mode 100644
index 0000000000..7a4a431f0a
--- /dev/null
+++ b/examples/acp-agent/cordis.snapshot-record.yml
@@ -0,0 +1,43 @@
+# Snapshot-test RECORD config: a real run whose persisted session JSONL is
+# harvested into a scenario fixture. Identical to cordis.yml (real llm-deepseek
+# adapter + JSONL persistence) — recording must exercise the REAL model so the
+# recorded log is a genuine product of the system. Needs DEEPSEEK_API_KEY.
+#
+# It is a separate file (rather than reusing cordis.yml) only so the snapshot
+# harness selects it explicitly via $DSH_SNAPSHOT=record and so its persistence
+# root can be pointed at the harness's harvest directory by the same env the
+# replay path uses. The graceful-shutdown path in start.ts flushes persistence
+# before exit so the harvested log is complete.
+
+- id: timer
+  name: '@cordisjs/plugin-timer'
+
+# Shared provider/tool core, INCLUDING the real llm-deepseek adapter.
+- id: base
+  name: '@cordisjs/plugin-include'
+  config:
+    path: '../base.yml'
+
+- id: agent-loop
+  name: '@deepseek-ai/dsh-agent-loop'
+  config:
+    agents: []
+
+- id: session-persistence
+  name: '@deepseek-ai/dsh-session-persistence-jsonl'
+  config:
+    root: !!js process.env.DSH_SNAPSHOT_SESSIONS_ROOT
+
+- id: acp
+  name: '@deepseek-ai/dsh-acp'
+  config:
+    model: deepseek-v4-flash
+    systemPrompt: |
+      You are a coding assistant driven over the Agent Client Protocol.
+
+      Your only tools are bash (plus bash_output/bash_kill for background
+      tasks). Do ALL file operations through bash: read with cat/sed/head,
+      search with grep, write with heredocs (cat <<'EOF' > file), edit with
+      sed or a rewrite. Each bash call runs in a fresh shell — pass workdir
+      instead of cd. Check the [exit code: N] marker; verify your work. Keep
+      answers brief and factual.
diff --git a/examples/acp-agent/cordis.snapshot.yml b/examples/acp-agent/cordis.snapshot.yml
new file mode 100644
index 0000000000..ed5690db18
--- /dev/null
+++ b/examples/acp-agent/cordis.snapshot.yml
@@ -0,0 +1,69 @@
+# Snapshot-test REPLAY config: the acp-agent plugin tree with the model replaced
+# by llm-replay (serves a recorded session JSONL — no API key, no network).
+#
+# This does NOT include ../base.yml: base.yml always loads
+# @deepseek-ai/dsh-llm-deepseek, whose apply() throws without DEEPSEEK_API_KEY,
+# so a keyless replay run would die at boot. We inline the providerless core
+# instead and install llm-replay where the adapter would be.
+#
+# stdout is reserved for the ACP JSON-RPC protocol — no stdout logger (see
+# cordis.yml). The replay fixture path comes from $DSH_SNAPSHOT_FILE (and an
+# optional $DSH_SNAPSHOT_OVERRIDE sidecar), set by the snapshot harness.
+
+- id: timer
+  name: '@cordisjs/plugin-timer'
+
+# Providerless core (everything base.yml has EXCEPT llm-deepseek).
+- id: llm
+  name: '@deepseek-ai/dsh-llm'
+
+- id: sessions
+  name: '@deepseek-ai/dsh-session'
+
+- id: system-prompt
+  name: '@deepseek-ai/dsh-system-prompt'
+
+- id: tools
+  name: '@deepseek-ai/dsh-tools'
+
+- id: agents
+  name: '@deepseek-ai/dsh-agent'
+
+- id: invariants
+  name: '@deepseek-ai/dsh-invariants'
+
+- id: bash
+  name: '@deepseek-ai/dsh-bash-local'
+  config:
+    timeoutMs: 60000
+
+- id: tool-bash
+  name: '@deepseek-ai/dsh-tool-bash'
+
+# The replay adapter: short-circuits llm/stream with the recorded log's chunks.
+- id: llm-replay
+  name: './src/llm-replay.ts'
+
+- id: agent-loop
+  name: '@deepseek-ai/dsh-agent-loop'
+  config:
+    agents: []
+
+- id: session-persistence
+  name: '@deepseek-ai/dsh-session-persistence-jsonl'
+  config:
+    root: !!js process.env.DSH_SNAPSHOT_SESSIONS_ROOT
+
+- id: acp
+  name: '@deepseek-ai/dsh-acp'
+  config:
+    model: deepseek-v4-flash
+    systemPrompt: |
+      You are a coding assistant driven over the Agent Client Protocol.
+
+      Your only tools are bash (plus bash_output/bash_kill for background
+      tasks). Do ALL file operations through bash: read with cat/sed/head,
+      search with grep, write with heredocs (cat <<'EOF' > file), edit with
+      sed or a rewrite. Each bash call runs in a fresh shell — pass workdir
+      instead of cd. Check the [exit code: N] marker; verify your work. Keep
+      answers brief and factual.
diff --git a/examples/acp-agent/start.ts b/examples/acp-agent/start.ts
index 1c97c0c8c0..01bcc32093 100644
--- a/examples/acp-agent/start.ts
+++ b/examples/acp-agent/start.ts
@@ -2,20 +2,36 @@ import { pathToFileURL } from 'node:url'
 import { Context } from 'cordis'
 import Loader from '@cordisjs/plugin-loader'
 
+// Snapshot-test modes (set by the snapshot harness via env):
+//   DSH_SNAPSHOT=replay  — load cordis.snapshot.yml (providerless; llm-replay
+//                          serves a recorded session log). Skip .env so a stray
+//                          key can never trigger a live model call.
+//   DSH_SNAPSHOT=record  — load cordis.snapshot-record.yml (the real adapter +
+//                          persistence) so a real run can be harvested.
+// Absent — the normal demo (cordis.yml), driven by a real editor.
+const snapshotMode = process.env.DSH_SNAPSHOT
+const configPath = snapshotMode === 'replay' ? './cordis.snapshot.yml'
+  : snapshotMode === 'record' ? './cordis.snapshot-record.yml'
+    : './cordis.yml'
+
 // Load DEEPSEEK_API_KEY / DEEPSEEK_BASE_URL from a gitignored repo-root .env
 // (Node native). Absent file is fine — the environment may already carry them.
+// In REPLAY mode we deliberately skip this: replay must never reach the network,
+// so we don't want a present .env to enable a live call.
 //
 // IMPORTANT: this server speaks ACP JSON-RPC on stdout. Do NOT add any
 // stdout logging here or in cordis.yml — it would corrupt the protocol frames.
 // A present-but-unreadable/malformed .env is a real misconfiguration: surface
 // it on STDERR (never stdout) rather than silently running with the wrong env.
-try {
-  process.loadEnvFile(new URL('../../.env', import.meta.url).pathname)
-} catch (error) {
-  if ((error as NodeJS.ErrnoException | null)?.code !== 'ENOENT') {
-    process.stderr.write(`acp-agent: failed to load .env: ${String(error)}\n`)
+if (snapshotMode !== 'replay') {
+  try {
+    process.loadEnvFile(new URL('../../.env', import.meta.url).pathname)
+  } catch (error) {
+    if ((error as NodeJS.ErrnoException | null)?.code !== 'ENOENT') {
+      process.stderr.write(`acp-agent: failed to load .env: ${String(error)}\n`)
+    }
+    // ENOENT (no .env) is fine — rely on the ambient environment.
   }
-  // ENOENT (no .env) is fine — rely on the ambient environment.
 }
 
 const ctx = new Context()
@@ -25,6 +41,18 @@ await ctx.plugin(Loader)
 await ctx.loader.create({
   name: '@cordisjs/plugin-include',
   config: {
-    path: './cordis.yml',
+    path: configPath,
   },
 })
+
+// Graceful shutdown for snapshot RECORD runs: when the client closes our stdin
+// (it is done driving the session), dispose the whole context. Disposal awaits
+// the agent-loop teardown and the persistence backend's final `session/flush`,
+// so the recorded `.jsonl` is fully written before the process exits and the
+// harness harvests it. (In a normal editor session stdin stays open for the
+// connection's lifetime; the editor kills the process, so this never fires.)
+if (snapshotMode !== undefined) {
+  process.stdin.on('end', () => {
+    void ctx.fiber.dispose().then(() => { process.exit(0) })
+  })
+}
diff --git a/examples/acp-agent/tests/acp.snapshot.ts b/examples/acp-agent/tests/acp.snapshot.ts
new file mode 100644
index 0000000000..a2422d4395
--- /dev/null
+++ b/examples/acp-agent/tests/acp.snapshot.ts
@@ -0,0 +1,90 @@
+import { readFile, readdir, writeFile } from 'node:fs/promises'
+import { existsSync } from 'node:fs'
+import { fileURLToPath } from 'node:url'
+import { dirname, join } from 'node:path'
+import { describe, expect, it } from 'vitest'
+import { type InputScript, runScenario } from './snapshot-harness.ts'
+import { type NormalizeContext, normalizeSessionLog, normalizeStdout } from './snapshot-normalize.ts'
+
+/**
+ * ACP snapshot tests (REPLAY by default, keyless). Each scenario under
+ * `snapshots/<name>/` ships an `input.json` (the client stdin script) and a
+ * recorded `session.jsonl` fixture; replay boots the real acp-agent subprocess,
+ * drives it, and diffs the normalized stdout transcript (and, for model
+ * scenarios, the re-persisted session log) against committed goldens.
+ *
+ * `pnpm run test:snapshot:record` (DSH_SNAPSHOT=record + -u) re-records the
+ * fixtures against the real API and refreshes the goldens in one pass.
+ */
+
+const SNAPSHOTS_DIR = join(dirname(fileURLToPath(import.meta.url)), 'snapshots')
+const RECORDING = process.env.DSH_SNAPSHOT === 'record'
+
+/** A scenario and whether it makes any model call (→ has a behavioral JSONL golden). */
+interface Scenario {
+  name: string
+  /** Whether the scenario drives at least one model turn (so a JSONL golden applies). */
+  hasModelTurn: boolean
+}
+
+const SCENARIOS: Scenario[] = [
+  { name: 'handshake', hasModelTurn: false },
+]
+
+for (const scenario of SCENARIOS) {
+  describe(`snapshot: ${scenario.name}`, () => {
+    it('matches the stdout transcript golden', async () => {
+      const dir = join(SNAPSHOTS_DIR, scenario.name)
+      const input = JSON.parse(await readFile(join(dir, 'input.json'), 'utf8')) as InputScript
+      const overrideFile = join(dir, 'replay.override.json')
+      const result = await runScenario(input, {
+        mode: RECORDING ? 'record' : 'replay',
+        fixtureFile: join(dir, 'session.jsonl'),
+        ...existsSync(overrideFile) ? { overrideFile } : {},
+      })
+
+      const ctx: NormalizeContext = {
+        sessionIds: result.sessionId !== undefined ? [result.sessionId] : [],
+        cwd: result.cwd,
+      }
+
+      // RECORD mode: persist the freshly-harvested log back to the scenario's
+      // session.jsonl fixture (a model scenario must produce one). `--update`
+      // refreshes the Vitest goldens but NOT this fixture, so write it here.
+      if (RECORDING && scenario.hasModelTurn) {
+        expect(result.sessionLog, 'record produced no session log to harvest').toBeDefined()
+        await writeFile(join(dir, 'session.jsonl'), result.sessionLog as string)
+      }
+
+      await expect(normalizeStdout(result.rawStdout, ctx))
+        .toMatchFileSnapshot(join(dir, 'stdout.golden.txt'))
+
+      if (scenario.hasModelTurn) {
+        expect(result.sessionLog, 'a model scenario must persist a session log').toBeDefined()
+        await expect(normalizeSessionLog(result.sessionLog as string, ctx))
+          .toMatchFileSnapshot(join(dir, 'session.golden.txt'))
+      }
+    })
+  })
+}
+
+describe('snapshot fixtures', () => {
+  it('every scenario directory is registered (no orphans)', async () => {
+    // toMatchFileSnapshot does not prune orphaned golden/fixture files, so a
+    // renamed/removed scenario could leave a stale dir that nothing exercises.
+    // Fail loud on any snapshots/<dir> not present in SCENARIOS.
+    const entries = await readdir(SNAPSHOTS_DIR, { withFileTypes: true })
+    const onDisk = entries.filter(e => e.isDirectory()).map(e => e.name).sort()
+    const registered = SCENARIOS.map(s => s.name).sort()
+    expect(onDisk).toEqual(registered)
+  })
+
+  it('every registered scenario has its required fixture files', async () => {
+    for (const { name } of SCENARIOS) {
+      const dir = join(SNAPSHOTS_DIR, name)
+      expect(existsSync(join(dir, 'input.json')), `${name}/input.json`).toBe(true)
+      expect(existsSync(join(dir, 'session.jsonl')), `${name}/session.jsonl`).toBe(true)
+      expect(existsSync(join(dir, 'stdout.golden.txt')), `${name}/stdout.golden.txt`).toBe(true)
+    }
+  })
+})
diff --git a/examples/acp-agent/tests/snapshot-harness.ts b/examples/acp-agent/tests/snapshot-harness.ts
new file mode 100644
index 0000000000..7fb7eec21d
--- /dev/null
+++ b/examples/acp-agent/tests/snapshot-harness.ts
@@ -0,0 +1,231 @@
+/**
+ * Shared harness for the ACP snapshot tests. A plain module (NOT a *.spec.ts /
+ * *.snapshot.ts) so importing it never re-registers another file's tests.
+ *
+ * It boots the REAL examples/acp-agent subprocess via the cordis Loader (so the
+ * export-shape bug class stays guarded — see docs/postmortem/0001), drives it
+ * over real ACP JSON-RPC stdio with a deterministic input script, tees raw
+ * stdout (for the golden + a purity check) into an SDK `ClientSideConnection`,
+ * and — in record mode — harvests the persisted session JSONL after a graceful
+ * shutdown flush. Two pure normalizers turn the captured stdout frames and the
+ * session-log events into stable, snapshot-able text.
+ *
+ * See docs/rfc/implemented/2026-06-19-acp-snapshot-tests.md.
+ */
+
+import { spawn, type ChildProcessWithoutNullStreams } from 'node:child_process'
+import { mkdtemp, readFile, readdir, rm } from 'node:fs/promises'
+import { tmpdir } from 'node:os'
+import { join } from 'node:path'
+import { fileURLToPath } from 'node:url'
+import { Readable, Writable } from 'node:stream'
+import {
+  ClientSideConnection,
+  ndJsonStream,
+  PROTOCOL_VERSION,
+  type Agent as AcpAgent,
+  type Client,
+  type RequestPermissionRequest,
+  type RequestPermissionResponse,
+  type SessionNotification,
+} from '@agentclientprotocol/sdk'
+
+const startScript = fileURLToPath(new URL('../start.ts', import.meta.url))
+const tsxLoader = fileURLToPath(import.meta.resolve('tsx'))
+// The repo-root tsconfig: dev/test run UNBUILT and the `@deepseek-ai/dsh-*`
+// imports resolve through its `paths` map. The child's cwd is a temp dir
+// OUTSIDE the repo, so tsx's upward search would miss it — point tsx at the
+// repo tsconfig explicitly (same fix the e2e harness uses). Repo root is four
+// levels up from this file (examples/acp-agent/tests).
+const repoTsconfig = fileURLToPath(new URL('../../../tsconfig.json', import.meta.url))
+
+/**
+ * One step of a scenario's deterministic input script (`input.json`). The
+ * harness interprets these in order. `newSession` captures the server-issued
+ * (random) session id into a `{{sessionId}}` variable that later steps
+ * reference, since a committed file cannot know the id in advance.
+ */
+type InputStep =
+  | { op: 'initialize'; terminalOutput?: boolean }
+  | { op: 'newSession' }
+  | { op: 'prompt'; text: string }
+  | { op: 'cancel' }
+
+/** A scenario's `input.json`: an ordered list of input steps. */
+export interface InputScript {
+  steps: InputStep[]
+}
+
+/** The result of running a scenario: raw stdout + the harvested session log. */
+export interface RunResult {
+  /** Raw stdout bytes (decoded utf8), every newline-delimited JSON-RPC frame. */
+  rawStdout: string
+  /** stderr (for diagnostics on failure). */
+  stderr: string
+  /** The session id the server issued (undefined if no session was created). */
+  sessionId?: string
+  /** The temp cwd the session ran in (the bash workspace). */
+  cwd: string
+  /** The persisted session log's content, if one was produced. */
+  sessionLog?: string
+}
+
+interface RunOptions {
+  /** `replay` (default, keyless) or `record` (real API, harvests the log). */
+  mode: 'replay' | 'record'
+  /** The recorded session JSONL fixture path (replay reads it; record writes near it). */
+  fixtureFile: string
+  /** Optional sidecar override path (replay). */
+  overrideFile?: string
+}
+
+/**
+ * Run a scenario end-to-end against a freshly-spawned subprocess. Owns the
+ * child and its temp dirs; always tears them down. Returns the captured stdout
+ * and (record mode) the harvested session-log path.
+ */
+export async function runScenario(input: InputScript, opts: RunOptions): Promise<RunResult> {
+  const cwd = await mkdtemp(join(tmpdir(), 'acp-snap-cwd-'))
+  const sessionsRoot = await mkdtemp(join(tmpdir(), 'acp-snap-sessions-'))
+  const env: NodeJS.ProcessEnv = {
+    ...process.env,
+    TSX_TSCONFIG_PATH: repoTsconfig,
+    DSH_SNAPSHOT: opts.mode,
+    DSH_SNAPSHOT_FILE: opts.fixtureFile,
+    DSH_SNAPSHOT_SESSIONS_ROOT: sessionsRoot,
+    ...opts.overrideFile !== undefined ? { DSH_SNAPSHOT_OVERRIDE: opts.overrideFile } : {},
+  }
+
+  const child: ChildProcessWithoutNullStreams = spawn(
+    process.execPath,
+    ['--import', tsxLoader, startScript],
+    { cwd, env, stdio: ['pipe', 'pipe', 'pipe'] },
+  )
+
+  const rawBuffers: Buffer[] = []
+  const stderrChunks: string[] = []
+  child.stderr.setEncoding('utf8')
+  child.stderr.on('data', (c: string) => stderrChunks.push(c))
+
+  // Tee raw stdout: accumulate the bytes for the golden + purity check, and ALSO
+  // feed the same bytes to the SDK client through a passthrough. Buffer the raw
+  // bytes (not per-chunk utf8 strings) and decode once at the end, so a
+  // multibyte sequence split across two 'data' events can't corrupt the golden.
+  const passthrough = new Readable({ read() {} })
+  child.stdout.on('data', (buf: Buffer) => {
+    rawBuffers.push(buf)
+    passthrough.push(buf)
+  })
+  child.stdout.on('end', () => passthrough.push(null))
+
+  const stream = ndJsonStream(
+    Writable.toWeb(child.stdin) as WritableStream<Uint8Array>,
+    Readable.toWeb(passthrough) as ReadableStream<Uint8Array>,
+  )
+  const makeClient = (_agent: AcpAgent): Client => ({
+    sessionUpdate(_params: SessionNotification): Promise<void> {
+      return Promise.resolve()
+    },
+    requestPermission(_params: RequestPermissionRequest): Promise<RequestPermissionResponse> {
+      return Promise.resolve({ outcome: { outcome: 'cancelled' } })
+    },
+  })
+  const client = new ClientSideConnection(makeClient, stream)
+
+  let sessionId: string | undefined
+  let sessionLog: string | undefined
+  try {
+    for (const step of input.steps) {
+      await runStep(client, step, cwd, () => sessionId, (id) => { sessionId = id })
+    }
+    // Done driving: close stdin so the server disposes gracefully (flushing
+    // persistence) and exits. Then await exit so the harvested log is complete.
+    child.stdin.end()
+    await waitForExit(child)
+    // Harvest the persisted log (if any) while the temp dirs still exist.
+    const sessionLogPath = await findSessionLog(sessionsRoot)
+    if (sessionLogPath !== undefined) sessionLog = await readFile(sessionLogPath, 'utf8')
+  } finally {
+    // Failure-safe teardown: kill a still-running child and drop the temp dirs
+    // even if a step/harvest threw, so a flaky run never leaks a process or dir.
+    if (child.exitCode === null && child.signalCode === null) {
+      child.kill('SIGKILL')
+      await waitForExit(child)
+    }
+    await rm(cwd, { recursive: true, force: true })
+    await rm(sessionsRoot, { recursive: true, force: true })
+  }
+
+  return {
+    rawStdout: Buffer.concat(rawBuffers).toString('utf8'),
+    stderr: stderrChunks.join(''),
+    cwd,
+    ...sessionId !== undefined ? { sessionId } : {},
+    ...sessionLog !== undefined ? { sessionLog } : {},
+  }
+}
+
+/** Drive one input step over the client connection. */
+async function runStep(
+  client: ClientSideConnection,
+  step: InputStep,
+  cwd: string,
+  getSessionId: () => string | undefined,
+  setSessionId: (id: string) => void,
+): Promise<void> {
+  switch (step.op) {
+    case 'initialize':
+      await client.initialize({
+        protocolVersion: PROTOCOL_VERSION,
+        clientCapabilities: step.terminalOutput === true ? { _meta: { terminal_output: true } } : {},
+      })
+      return
+    case 'newSession': {
+      const { sessionId } = await client.newSession({ cwd, mcpServers: [] })
+      setSessionId(sessionId)
+      return
+    }
+    case 'prompt': {
+      const sessionId = getSessionId()
+      if (sessionId === undefined) throw new Error('snapshot-harness: prompt before newSession')
+      await client.prompt({ sessionId, prompt: [{ type: 'text', text: step.text }] })
+      return
+    }
+    case 'cancel': {
+      const sessionId = getSessionId()
+      if (sessionId === undefined) throw new Error('snapshot-harness: cancel before newSession')
+      await client.cancel({ sessionId })
+      return
+    }
+    default:
+      throw new Error(`snapshot-harness: unknown input op ${JSON.stringify(step)}`)
+  }
+}
+
+/** Resolve once the child process exits (any code/signal). */
+function waitForExit(child: ChildProcessWithoutNullStreams): Promise<void> {
+  if (child.exitCode !== null || child.signalCode !== null) return Promise.resolve()
+  return new Promise<void>(resolve => child.once('exit', () => { resolve() }))
+}
+
+/** Find the single produced `.jsonl` session log under a sessions root, if any. */
+async function findSessionLog(root: string): Promise<string | undefined> {
+  let cwdDirs: string[]
+  try {
+    cwdDirs = await readdir(root)
+  } catch {
+    return undefined
+  }
+  for (const dir of cwdDirs) {
+    const sub = join(root, dir)
+    let files: string[]
+    try {
+      files = await readdir(sub)
+    } catch {
+      continue
+    }
+    const jsonl = files.find(f => f.endsWith('.jsonl'))
+    if (jsonl !== undefined) return join(sub, jsonl)
+  }
+  return undefined
+}
diff --git a/examples/acp-agent/tests/snapshot-normalize.spec.ts b/examples/acp-agent/tests/snapshot-normalize.spec.ts
new file mode 100644
index 0000000000..339cb8493f
--- /dev/null
+++ b/examples/acp-agent/tests/snapshot-normalize.spec.ts
@@ -0,0 +1,93 @@
+import { describe, expect, it } from 'vitest'
+import { type NormalizeContext, normalizeSessionLog, normalizeStdout } from '../tests/snapshot-normalize.ts'
+
+/**
+ * Unit tests for the pure snapshot normalizers. Live as a *.spec.ts (runs in
+ * the default unit gate) and import the harness-side normalizers directly.
+ */
+
+const ctx: NormalizeContext = {
+  sessionIds: ['11111111-2222-3333-4444-555555555555'],
+  cwd: '/tmp/acp-snap-cwd-abc123',
+}
+
+describe('normalizeStdout', () => {
+  it('rewrites JSON-RPC ids to a stable first-seen sequence', () => {
+    const raw = [
+      JSON.stringify({ jsonrpc: '2.0', id: 42, method: 'initialize' }),
+      JSON.stringify({ jsonrpc: '2.0', id: 42, result: {} }),
+      JSON.stringify({ jsonrpc: '2.0', id: 99, method: 'session/new' }),
+    ].join('\n')
+    const out = normalizeStdout(raw, ctx)
+    expect(out).toContain('"id": 1')
+    expect(out).toContain('"id": 2')
+    expect(out).not.toContain('42')
+    expect(out).not.toContain('99')
+  })
+
+  it('scrubs the cwd and session id anywhere they appear', () => {
+    const raw = JSON.stringify({
+      jsonrpc: '2.0', method: 'session/update',
+      params: { sessionId: ctx.sessionIds[0], cwd: ctx.cwd, note: `at ${ctx.cwd}/x` },
+    })
+    const out = normalizeStdout(raw, ctx)
+    expect(out).toContain('{{sessionId}}')
+    expect(out).toContain('{{cwd}}')
+    expect(out).not.toContain(ctx.cwd)
+    expect(out).not.toContain(ctx.sessionIds[0] as string)
+  })
+
+  it('scrubs a stray UUID not in the known list', () => {
+    const raw = JSON.stringify({ jsonrpc: '2.0', method: 'x', params: { id: 'aaaaaaaa-bbbb-cccc-dddd-eeeeeeeeeeee' } })
+    expect(normalizeStdout(raw, ctx)).toContain('{{sessionId}}')
+  })
+
+  it('leaves notification frames without an id untouched in id-space', () => {
+    const raw = JSON.stringify({ jsonrpc: '2.0', method: 'session/update', params: {} })
+    const out = normalizeStdout(raw, ctx)
+    expect(out).not.toContain('"id"')
+  })
+
+  it('throws on a non-JSON stdout line (the purity check)', () => {
+    const raw = `${JSON.stringify({ jsonrpc: '2.0', id: 1 })}\noops a log leaked\n`
+    expect(() => normalizeStdout(raw, ctx)).toThrow()
+  })
+
+  it('ignores blank lines', () => {
+    const raw = `\n${JSON.stringify({ jsonrpc: '2.0', id: 1, method: 'm' })}\n\n`
+    expect(() => normalizeStdout(raw, ctx)).not.toThrow()
+  })
+})
+
+describe('normalizeSessionLog', () => {
+  const header = (over: object) => JSON.stringify({ type: 'session', version: 1, id: 's', createdAt: 123, ...over })
+  const event = (over: object) => JSON.stringify({ type: 'turn/start', seq: 1, time: 999, data: { turn: 1 }, ...over })
+
+  it('zeroes the header createdAt', () => {
+    const out = normalizeSessionLog(`${header({})}\n`, ctx)
+    expect(out).toContain('"createdAt": 0')
+    expect(out).not.toContain('123')
+  })
+
+  it('zeroes each event time but keeps seq', () => {
+    const out = normalizeSessionLog(`${header({})}\n${event({ seq: 7, time: 999 })}\n`, ctx)
+    expect(out).toContain('"time": 0')
+    expect(out).toContain('"seq": 7') // seq is deterministic — NOT scrubbed
+    expect(out).not.toContain('999')
+  })
+
+  it('scrubs cwd and session id deep inside event data', () => {
+    const ev = JSON.stringify({
+      type: 'tool/result', seq: 2, time: 5,
+      data: { content: [{ type: 'text', text: `wrote ${ctx.cwd}/proof.txt` }] },
+    })
+    const out = normalizeSessionLog(`${header({ cwd: ctx.cwd })}\n${ev}\n`, ctx)
+    expect(out).toContain('{{cwd}}')
+    expect(out).not.toContain(ctx.cwd)
+  })
+
+  it('scrubs the session id in the header', () => {
+    const out = normalizeSessionLog(`${header({ id: ctx.sessionIds[0] })}\n`, ctx)
+    expect(out).toContain('{{sessionId}}')
+  })
+})
diff --git a/examples/acp-agent/tests/snapshot-normalize.ts b/examples/acp-agent/tests/snapshot-normalize.ts
new file mode 100644
index 0000000000..c729c429b8
--- /dev/null
+++ b/examples/acp-agent/tests/snapshot-normalize.ts
@@ -0,0 +1,102 @@
+/**
+ * Pure normalizers for the ACP snapshot goldens. They replace the
+ * non-deterministic values in the two captured surfaces — the stdout JSON-RPC
+ * transcript and the persisted session JSONL — with stable tokens, so a golden
+ * compare reflects behavior, not run-to-run noise. Kept dependency-free and
+ * side-effect-free so they unit-test trivially.
+ *
+ * Scrubbed: `randomUUID()` session ids → `{{sessionId}}`; the temp `mkdtemp`
+ * cwd → `{{cwd}}` (it appears in terminal-card `_meta` and the log header);
+ * JSON-RPC request `id` → a stable per-transcript sequence; the log's per-event
+ * `time` (epoch ms) and header `createdAt` → 0. NOT scrubbed: the log's `seq`
+ * (deterministic — `seq = log.length`, part of the event-log contract).
+ *
+ * See docs/rfc/implemented/2026-06-19-acp-snapshot-tests.md.
+ */
+
+const SESSION_ID = '{{sessionId}}'
+const CWD = '{{cwd}}'
+
+/** A UUID v4 string, the shape `randomUUID()` produces for session ids. */
+const UUID_RE = /[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}/gi
+
+/** Inputs the normalizers need to recognize a run's volatile values. */
+export interface NormalizeContext {
+  /** The session id(s) the run issued — replaced with `{{sessionId}}`. */
+  sessionIds: string[]
+  /** The temp cwd the run used — replaced with `{{cwd}}`. */
+  cwd: string
+}
+
+/** Replace cwd, session ids, and any stray UUID with stable tokens in a string. */
+function scrubString(value: string, ctx: NormalizeContext): string {
+  let out = value
+  // cwd first (longest, most specific), then explicit session ids, then any
+  // residual UUID (covers ids that appear in places we didn't enumerate).
+  out = out.split(ctx.cwd).join(CWD)
+  for (const id of ctx.sessionIds) out = out.split(id).join(SESSION_ID)
+  out = out.replace(UUID_RE, SESSION_ID)
+  return out
+}
+
+/** Recursively scrub a parsed JSON value (strings replaced; structure kept). */
+function scrubValue(value: unknown, ctx: NormalizeContext): unknown {
+  if (typeof value === 'string') return scrubString(value, ctx)
+  if (Array.isArray(value)) return value.map(v => scrubValue(v, ctx))
+  if (value !== null && typeof value === 'object') {
+    const out: Record<string, unknown> = {}
+    for (const [k, v] of Object.entries(value)) out[k] = scrubValue(v, ctx)
+    return out
+  }
+  return value
+}
+
+/**
+ * Normalize a raw stdout transcript (newline-delimited JSON-RPC frames) into a
+ * stable, line-diffable golden: one pretty-printed frame per block, with the
+ * JSON-RPC `id` rewritten to a per-transcript sequence (1, 2, 3, …) and all
+ * volatile strings scrubbed. Throws if any non-empty line is not valid JSON —
+ * that doubles as the stdout-purity check (no logger leaked onto the protocol).
+ */
+export function normalizeStdout(rawStdout: string, ctx: NormalizeContext): string {
+  const lines = rawStdout.split('\n').filter(line => line.trim().length > 0)
+  // Map each distinct JSON-RPC id (request/response correlate by id) to a stable
+  // sequence number, in first-seen order, so id churn doesn't perturb the golden.
+  const idSeq = new Map<string, number>()
+  const stableId = (id: unknown): number => {
+    const key = JSON.stringify(id)
+    let n = idSeq.get(key)
+    if (n === undefined) { n = idSeq.size + 1; idSeq.set(key, n) }
+    return n
+  }
+  const frames = lines.map((line) => {
+    const frame = JSON.parse(line) as Record<string, unknown>
+    if ('id' in frame && frame.id !== undefined && frame.id !== null) {
+      frame.id = stableId(frame.id)
+    }
+    return scrubValue(frame, ctx) as Record<string, unknown>
+  })
+  return frames.map(f => JSON.stringify(f, null, 2)).join('\n') + '\n'
+}
+
+/**
+ * Normalize a session JSONL log into a stable golden: the header line's
+ * volatile fields (`createdAt`, `id`, `cwd`) and every event's `time` are
+ * zeroed/scrubbed, all volatile strings scrubbed, and `seq` is LEFT INTACT
+ * (deterministic by contract). One pretty-printed record per block.
+ */
+export function normalizeSessionLog(rawLog: string, ctx: NormalizeContext): string {
+  const lines = rawLog.split('\n').filter(line => line.trim().length > 0)
+  const records = lines.map((line) => {
+    const record = JSON.parse(line) as Record<string, unknown>
+    // Header line: { type: 'session', createdAt, id, cwd, … }.
+    if (record.type === 'session') {
+      if ('createdAt' in record) record.createdAt = 0
+    } else if ('time' in record) {
+      // Event line: zero the epoch-ms timestamp; keep seq (deterministic).
+      record.time = 0
+    }
+    return scrubValue(record, ctx) as Record<string, unknown>
+  })
+  return records.map(r => JSON.stringify(r, null, 2)).join('\n') + '\n'
+}
diff --git a/examples/acp-agent/tests/snapshots/handshake/input.json b/examples/acp-agent/tests/snapshots/handshake/input.json
new file mode 100644
index 0000000000..e1e84be919
--- /dev/null
+++ b/examples/acp-agent/tests/snapshots/handshake/input.json
@@ -0,0 +1,6 @@
+{
+  "steps": [
+    { "op": "initialize" },
+    { "op": "newSession" }
+  ]
+}
diff --git a/examples/acp-agent/tests/snapshots/handshake/session.jsonl b/examples/acp-agent/tests/snapshots/handshake/session.jsonl
new file mode 100644
index 0000000000..ab44090be6
--- /dev/null
+++ b/examples/acp-agent/tests/snapshots/handshake/session.jsonl
@@ -0,0 +1 @@
+{"type":"session","version":1,"id":"00000000-0000-0000-0000-000000000000","createdAt":0}
diff --git a/examples/acp-agent/tests/snapshots/handshake/stdout.golden.txt b/examples/acp-agent/tests/snapshots/handshake/stdout.golden.txt
new file mode 100644
index 0000000000..ec34bb7f4a
--- /dev/null
+++ b/examples/acp-agent/tests/snapshots/handshake/stdout.golden.txt
@@ -0,0 +1,27 @@
+{
+  "jsonrpc": "2.0",
+  "id": 1,
+  "result": {
+    "protocolVersion": 1,
+    "agentInfo": {
+      "name": "deepseek-harness-acp",
+      "version": "0.0.1"
+    },
+    "agentCapabilities": {
+      "loadSession": true,
+      "promptCapabilities": {
+        "image": false,
+        "audio": false,
+        "embeddedContext": false
+      }
+    },
+    "authMethods": []
+  }
+}
+{
+  "jsonrpc": "2.0",
+  "id": 2,
+  "result": {
+    "sessionId": "{{sessionId}}"
+  }
+}
diff --git a/knip.json b/knip.json
index 0693f71c02..cb836485f1 100644
--- a/knip.json
+++ b/knip.json
@@ -8,7 +8,8 @@
         "examples/echo-agent/src/*.ts",
         "examples/coding-agent/src/*.ts",
         "examples/acp-agent/src/*.ts",
-        "examples/acp-agent/tests/**/*.e2e.ts"
+        "examples/acp-agent/tests/**/*.e2e.ts",
+        "examples/acp-agent/tests/**/*.snapshot.ts"
       ],
       "project": ["scripts/**/*.ts", "examples/**/*.ts"]
     },
diff --git a/lefthook.yml b/lefthook.yml
index 24600e4985..2a255424fa 100644
--- a/lefthook.yml
+++ b/lefthook.yml
@@ -25,6 +25,9 @@ pre-push:
     - name: test
       run: pnpm run test
 
+    - name: snapshot
+      run: pnpm run test:snapshot
+
     - name: hygiene
       run: pnpm run hygiene
 
diff --git a/package.json b/package.json
index fc8bce92c4..b892288a0e 100644
--- a/package.json
+++ b/package.json
@@ -19,6 +19,8 @@
     "test": "vitest run",
     "test:coverage": "vitest run --coverage",
     "test:e2e": "vitest run --config vitest.e2e.config.ts",
+    "test:snapshot": "vitest run --config vitest.snapshot.config.ts",
+    "test:snapshot:record": "DSH_SNAPSHOT=record vitest run --config vitest.snapshot.config.ts --update",
     "knip": "knip",
     "publint": "tsx scripts/publint-all.ts",
     "doc-typecheck": "tsx scripts/doc-typecheck.ts",
diff --git a/vitest.snapshot.config.ts b/vitest.snapshot.config.ts
new file mode 100644
index 0000000000..1fb7ee48a4
--- /dev/null
+++ b/vitest.snapshot.config.ts
@@ -0,0 +1,32 @@
+import tsconfigPaths from 'vite-tsconfig-paths'
+import { defineConfig } from 'vitest/config'
+
+// Snapshot tests: `pnpm run test:snapshot`, file pattern *.snapshot.ts.
+// REPLAY by default — they boot the real acp-agent subprocess against a
+// recorded session JSONL fixture (no API key, no network) and diff the
+// normalized stdout transcript + re-persisted log against committed goldens.
+// `pnpm run test:snapshot:record` (DSH_SNAPSHOT=record + -u) re-records the
+// fixtures against the real API and refreshes the goldens.
+//
+// Replay loads no .env; record reads DEEPSEEK_API_KEY from the env or a
+// gitignored repo-root .env (loaded here, mirroring the e2e config), so a
+// contributor with a key only in .env can still record.
+try {
+  process.loadEnvFile(new URL('.env', import.meta.url).pathname)
+} catch {
+  // No .env — fine; replay needs no key and record reads it from the env.
+}
+
+export default defineConfig({
+  // Same resolution note as vitest.config.ts: bare workspace names resolve
+  // through the root tsconfig paths map; the native option cannot do this.
+  plugins: [tsconfigPaths({ projects: ['./tsconfig.test.json'] })],
+  test: {
+    include: ['examples/*/tests/**/*.snapshot.ts'],
+    // Each test boots a subprocess; give it room, and run files one at a time
+    // (a record run hits the live API, and replay subprocess boot is heavy).
+    testTimeout: 120_000,
+    hookTimeout: 30_000,
+    fileParallelism: false,
+  },
+})