@knpkv/jira-clockify 1.3.0 → 1.4.0

package/dist/src/testing/fakeHeadless.js

@@ -0,0 +1,1149 @@
+/**
+ * `FakeHeadlessLayer` — the same service shape as the CLI's `HeadlessLayer`, with every external
+ * boundary faked and every write captured.
+ *
+ * **Why this exists**
+ *
+ * `HeadlessLayer` is fully live, which is why `commandSurface.test.ts` can only assert `--help`
+ * outcomes. This layer keeps the *real* service implementations (TimerService, ReconcileService,
+ * AgentSessionReader) and fakes only what leaves the process: Clockify, Jira, both auth services,
+ * config, the filesystem, the terminal, and the Coding Agent. Commands can then be run end to end
+ * and asserted on what they printed, proposed, and wrote.
+ *
+ * **How to read a test written against it**
+ *
+ * - `world.createdClockifyEntries` / `world.jiraWorklogs` are the writes that actually happened.
+ * - `world.describeRequests` / `world.describeBatches` do the same for the notes written onto
+ *   entries, so what a run *spends* on descriptions is assertable too.
+ * - `world.attributorRequests` is every session asked about and `world.attributorBatches` is the
+ *   number of *calls* those took, so both "cost nothing" and "batched, not one-by-one" are
+ *   assertable. `world.maxAttributorInFlight` is the most calls that ever overlapped.
+ * - `world.stdout` / `world.stderr` are the lines the command printed, split by stream, which is
+ *   what the JSON Output Contract is about.
+ * - `world.prompts` is what the terminal rendered. Prompt text goes through `Terminal.display`,
+ *   not `Console`, so this is the only place a confirmation's wording is observable.
+ * - `world.transcriptReads` is every transcript path opened, so a scope rule that promises a file is
+ *   never read can be asserted rather than assumed.
+ * - `world.transcripts` is the transcripts on disk, read live. Append to it mid-run to model a
+ *   session that starts, or grows, while a long-running command is watching.
+ * - `keep` scripts the row picker: `keep[i] === false` unchecks row `i`. Omitting `keep` entirely
+ *   sends no input at all, which is the same signal as a missing TTY.
+ *
+ * **Why it ships**
+ *
+ * It lives in `src` and is exported as `@knpkv/jira-clockify/testing.js` because it is the highest
+ * seam in the package and the only one a *consumer* can drive: a web server over these services has
+ * to be testable against captured writes rather than a real Clockify account. Kept out of the barrel
+ * so nothing production imports it by accident.
+ */
+import { NodePath } from "@effect/platform-node";
+import { ClockifyApiClient } from "@knpkv/clockify-api-client";
+import { JiraApiClient, JiraApiConfig } from "@knpkv/jira-api-client";
+import { JiraAuth } from "@knpkv/jira-cli/JiraAuth";
+import * as ByteSize from "effect/ByteSize";
+import * as Console from "effect/Console";
+import * as Effect from "effect/Effect";
+import * as FileSystem from "effect/FileSystem";
+import { HttpClient, HttpClientError, HttpClientRequest, HttpClientResponse } from "effect/http";
+import * as Layer from "effect/Layer";
+import * as Option from "effect/Option";
+import { systemError } from "effect/PlatformError";
+import * as Predicate from "effect/Predicate";
+import * as ChildProcessSpawner from "effect/process/ChildProcessSpawner";
+import * as Queue from "effect/Queue";
+import * as Redacted from "effect/Redacted";
+import * as Schema from "effect/Schema";
+import * as Stdio from "effect/Stdio";
+import * as Terminal from "effect/Terminal";
+import { defaultSessionAgentSettings } from "../agent/agentSettings.js";
+import { LogToStderrLive } from "../cli/layers.js";
+import { layer as agentSessionReaderLayer } from "../services/AgentSessionReader.js";
+import { ClockifyAuth } from "../services/ClockifyAuth.js";
+import { ConfigService } from "../services/ConfigService.js";
+import { HomeDirectory } from "../services/HomeDirectory.js";
+import { layer as issueFactsLayer } from "../services/IssueFacts.js";
+import { layer as reconcileServiceLayer } from "../services/ReconcileService.js";
+import { layer as savedEntriesLayer } from "../services/SavedEntries.js";
+import { SessionAttributor, SessionAttributorError } from "../services/SessionAttributor.js";
+import { layer as sourceLedgerLayer } from "../services/SourceLedger.js";
+import { StateWriter } from "../services/StateWriter.js";
+import { layer as ticketServiceLayer } from "../services/TicketService.js";
+import { layer as timerServiceLayer } from "../services/TimerService.js";
+export const FAKE_HOME = "/fake-home";
+export const FAKE_WORKSPACE_ID = "ws-fake";
+export const FAKE_USER_ID = "user-fake";
+/**
+ * The text of a Jira worklog comment, read back out of the document format Jira takes.
+ *
+ * Deliberately tolerant: a test asserts on what a reader would see, and walking the node tree for
+ * every `text` leaf says that without also pinning the paragraph structure around it.
+ */
+/**
+ * One node of Jira's document format, reduced to what reading a comment back needs.
+ *
+ * Decoded rather than inspected: the tree is external data, and a schema is what turns "some
+ * object" into something with a `text` and a `content` this can rely on.
+ */
+const AdfNode = Schema.Struct({
+    text: Schema.optional(Schema.String),
+    content: Schema.optional(Schema.Unknown)
+});
+const decodeAdfNode = Schema.decodeUnknownOption(AdfNode);
+/** The worklog fields the fake reads back off a POST body. */
+const WorklogPayload = Schema.Struct({
+    started: Schema.optional(Schema.String),
+    timeSpentSeconds: Schema.optional(Schema.Finite),
+    comment: Schema.optionalKey(Schema.Json)
+});
+const decodeWorklogPayload = Schema.decodeUnknownOption(Schema.fromJsonString(WorklogPayload));
+const commentText = (comment) => {
+    const texts = [];
+    const walk = (node) => {
+        if (Array.isArray(node)) {
+            for (const child of node)
+                walk(child);
+            return;
+        }
+        const decoded = decodeAdfNode(node);
+        if (Option.isNone(decoded))
+            return;
+        const { content, text } = decoded.value;
+        if (text !== undefined)
+            texts.push(text);
+        if (content !== undefined)
+            walk(content);
+    };
+    walk(comment);
+    return texts.join(" ");
+};
+/** Whoever the fake is logged in as unless a test says otherwise. */
+export const FAKE_ACCOUNT_ID = "acct-me";
+const defaultConfig = {
+    sessionAgent: defaultSessionAgentSettings,
+    defaultJql: "",
+    refreshInterval: 30,
+    projectMap: {},
+    workspaceId: null,
+    defaultProjectId: null,
+    defaultProjectName: null,
+    defaultBillable: true,
+    sessionRoots: [],
+    sessionTicketMap: {},
+    sessionIdleCapSeconds: 300,
+    sessionConfidenceFloor: 0.7,
+    sessionOwnership: "assigned",
+    sessionOwnershipOverrides: [],
+    sessionDwellSeconds: 900
+};
+const makeTimeEntry = (entry, id) => ({
+    id: entry.id ?? id,
+    description: entry.description,
+    billable: entry.billable ?? true,
+    userId: entry.userId ?? FAKE_USER_ID,
+    ...(entry.projectId !== undefined && { projectId: entry.projectId }),
+    ...(entry.taskId !== undefined && { taskId: entry.taskId }),
+    workspaceId: FAKE_WORKSPACE_ID,
+    timeInterval: { start: entry.start, ...((entry.end !== undefined) && { end: entry.end }) },
+    tagIds: entry.tagIds ?? [],
+    type: "REGULAR",
+    isLocked: false
+});
+/** A `File.Info` with no mtime, so the reader's mtime pre-filter never skips a fixture. */
+const fileInfo = (type) => ({
+    type,
+    mtime: Option.none(),
+    atime: Option.none(),
+    birthtime: Option.none(),
+    dev: 0,
+    ino: Option.none(),
+    mode: 0,
+    nlink: Option.none(),
+    uid: Option.none(),
+    gid: Option.none(),
+    rdev: Option.none(),
+    size: ByteSize.bytes(0),
+    blksize: Option.none(),
+    blocks: Option.none()
+});
+const TRANSCRIPT_ROOT = `${FAKE_HOME}/.claude/projects`;
+/**
+ * A read-only in-memory filesystem serving the transcript fixtures a test declared.
+ *
+ * Read live from `world.transcripts` on every call rather than snapshotted at layer construction.
+ * A watch is a command that runs while the transcripts underneath it grow, so a test has to be able
+ * to append a prompt between two ticks — with a snapshot, every tick would see the same day.
+ */
+const fakeFileSystemLayer = (transcripts, reads, written, unwritable, unreadable, afterFileWrite, pidNamespaceInode) => {
+    /**
+     * The project directory a transcript really lives in: the Claude CLI derives it from the working
+     * directory the session ran in, not from anything the test chose.
+     *
+     * Derived here rather than taken from the fixture key so a test cannot accidentally file an
+     * out-of-scope session inside a Session Root — which is the behaviour the reader's pre-read scope
+     * filter depends on. The key's own directory is the fallback for a fixture with no readable `cwd`.
+     */
+    const projectDirOf = (key, content) => {
+        for (const line of content.split("\n")) {
+            if (!line.includes("\"cwd\""))
+                continue;
+            const parsed = JSON.parse(line);
+            const cwd = Predicate.isObject(parsed) ? parsed["cwd"] : undefined;
+            if (Predicate.isString(cwd))
+                return cwd.replace(/[/.]/g, "-");
+        }
+        return key.split("/")[0] ?? key;
+    };
+    const fileName = (key) => key.slice(key.lastIndexOf("/") + 1);
+    const path = (key, content) => `${TRANSCRIPT_ROOT}/${projectDirOf(key, content)}/${fileName(key)}`;
+    const files = () => new Map(Object.entries(transcripts).map(([key, c]) => [path(key, c), c]));
+    const projectDirs = () => new Set(Object.entries(transcripts).map(([key, c]) => projectDirOf(key, c)));
+    const directories = () => {
+        const dirs = new Set([FAKE_HOME, `${FAKE_HOME}/.claude`, TRANSCRIPT_ROOT, `${FAKE_HOME}/.jcf`]);
+        for (const dir of projectDirs())
+            dirs.add(`${TRANSCRIPT_ROOT}/${dir}`);
+        return dirs;
+    };
+    const notFound = (method, path) => Effect.fail(systemError({
+        _tag: "NotFound",
+        module: "FileSystem",
+        method,
+        description: "No such file or directory",
+        pathOrDescriptor: path
+    }));
+    const permissionDenied = (method, path) => Effect.fail(systemError({
+        _tag: "PermissionDenied",
+        module: "FileSystem",
+        method,
+        description: "permission denied",
+        pathOrDescriptor: path
+    }));
+    const syncedHandle = (type) => ({
+        [FileSystem.FileTypeId]: FileSystem.FileTypeId,
+        stat: Effect.succeed(fileInfo(type)),
+        seek: () => Effect.succeed(ByteSize.bytes(0)),
+        sync: Effect.void,
+        read: () => Effect.succeed(0),
+        readAlloc: () => Effect.succeed(Option.none()),
+        truncate: () => Effect.void,
+        write: (buffer) => Effect.succeed(buffer.length),
+        writeAll: () => Effect.void
+    });
+    return FileSystem.layerNoop({
+        realPath: (path) => directories().has(path) || files().has(path) || written[path] !== undefined
+            ? Effect.succeed(path)
+            : notFound("realPath", path),
+        open: (path) => directories().has(path)
+            ? Effect.succeed(syncedHandle("Directory"))
+            : written[path] !== undefined
+                ? Effect.succeed(syncedHandle("File"))
+                : notFound("open", path),
+        exists: (path) => Effect.succeed(directories().has(path) || files().has(path) || written[path] !== undefined),
+        stat: (path) => unreadable.includes(path)
+            ? permissionDenied("stat", path)
+            : path === "/proc/self/ns/pid" && pidNamespaceInode !== undefined
+                ? Effect.succeed({ ...fileInfo("File"), ino: Option.some(pidNamespaceInode) })
+                : directories().has(path)
+                    ? Effect.succeed(fileInfo("Directory"))
+                    : files().has(path) || written[path] !== undefined
+                        ? Effect.succeed(fileInfo("File"))
+                        : notFound("stat", path),
+        readDirectory: (path) => {
+            if (unreadable.some((suffix) => path.endsWith(suffix)))
+                return permissionDenied("readDirectory", path);
+            if (path === TRANSCRIPT_ROOT)
+                return Effect.succeed([...projectDirs()]);
+            const prefix = `${path}/`;
+            const entries = [...files().keys()]
+                .filter((file) => file.startsWith(prefix))
+                .map((file) => file.slice(prefix.length));
+            return directories().has(path) ? Effect.succeed(entries) : notFound("readDirectory", path);
+        },
+        readFileString: (path) => {
+            if (path.startsWith(`${TRANSCRIPT_ROOT}/`))
+                reads.push(path);
+            // Matched by suffix, so a test names the fixture it wrote rather than reproducing the project
+            // directory encoding the Claude CLI derives from a session's `cwd`.
+            if (unreadable.some((suffix) => path.endsWith(suffix))) {
+                return permissionDenied("readFileString", path);
+            }
+            const stored = written[path];
+            if (stored !== undefined)
+                return Effect.succeed(stored);
+            const content = files().get(path);
+            return content === undefined ? notFound("readFileString", path) : Effect.succeed(content);
+        },
+        // The config service writes through this; nothing in these tests reads it back from disk.
+        makeDirectory: () => Effect.void,
+        // A real store, so anything the CLI writes and reads back — the watch lease — behaves. `wx` is
+        // honoured, because an exclusive create is exactly what the lease relies on for its exclusion.
+        writeFileString: (path, content, options) => unwritable.includes(path)
+            // Deliberately not `AlreadyExists`: telling those two apart is the whole point of the
+            // lease's fail-closed rule, so the fake must be able to produce the other one.
+            ? Effect.fail(systemError({
+                _tag: "PermissionDenied",
+                module: "FileSystem",
+                method: "writeFileString",
+                description: "permission denied",
+                pathOrDescriptor: path
+            }))
+            : options?.flag === "wx" && written[path] !== undefined
+                ? Effect.fail(systemError({
+                    _tag: "AlreadyExists",
+                    module: "FileSystem",
+                    method: "writeFileString",
+                    description: "file already exists",
+                    pathOrDescriptor: path
+                }))
+                : Effect.sync(() => {
+                    written[path] = content;
+                }).pipe(Effect.andThen(afterFileWrite(path))),
+        remove: (path) => Effect.sync(() => {
+            delete written[path];
+        }),
+        rename: (from, to) => written[from] === undefined
+            ? notFound("rename", from)
+            : unwritable.includes(to)
+                ? permissionDenied("rename", to)
+                : Effect.sync(() => {
+                    written[to] = written[from];
+                    delete written[from];
+                })
+    });
+};
+const keypress = (name) => ({
+    input: Option.none(),
+    key: { name, ctrl: false, meta: false, shift: false }
+});
+/**
+ * Translate "keep these rows" into the keys a multi-select actually receives.
+ *
+ * The picker's cursor starts on its two meta options (select-all, invert), so choice `i` sits at
+ * cursor position `i + 2`. Every row starts checked, so only the rows to *drop* need a `space`.
+ */
+export const pickerKeys = (keep) => {
+    const keys = [];
+    let cursor = 0;
+    keep.forEach((keepRow, index) => {
+        if (keepRow)
+            return;
+        const target = index + 2;
+        for (; cursor < target; cursor++)
+            keys.push(keypress("down"));
+        keys.push(keypress("space"));
+    });
+    keys.push(keypress("enter"));
+    return keys;
+};
+/** The conventional terminal width, and the narrowest the picker lays rows out for. */
+const DEFAULT_COLUMNS = 80;
+/**
+ * A terminal that replays scripted keypresses. Once the script is exhausted the input queue ends,
+ * which `Prompt` reports as `QuitError` — the same outcome as running without a TTY.
+ */
+const fakeTerminalLayer = (options, world) => {
+    let prompt = 0;
+    return Layer.succeed(Terminal.Terminal, Terminal.make({
+        columns: Effect.succeed(options.columns),
+        rows: Effect.succeed(24),
+        // Scoped in Effect 4, and the queue ends with `Cause.Done` rather than a terminal-specific
+        // error. An exhausted script is still what a missing TTY looks like to `Prompt`.
+        readInput: Effect.gen(function* () {
+            yield* Effect.sync(() => options.beforePromptInput?.(world));
+            const queue = yield* Queue.unbounded();
+            const promptInput = options.promptInputs?.[prompt++];
+            if (promptInput !== undefined)
+                yield* Queue.offerAll(queue, promptInput.map(keypress));
+            // No script at all means no input, so the prompt ends unanswered.
+            if (promptInput === undefined && options.keep !== undefined) {
+                yield* Queue.offerAll(queue, pickerKeys(options.keep));
+            }
+            yield* Queue.end(queue);
+            return queue;
+        }),
+        readLine: Effect.succeed(""),
+        display: (text) => Effect.sync(() => {
+            world.prompts.push(text);
+        })
+    }));
+};
+/** A console that keeps stdout and stderr apart, which is what the JSON Output Contract needs. */
+const captureConsoleLayer = (world, afterLog) => Layer.succeed(Console.Console, {
+    log: (...args) => {
+        const line = args.map(String).join(" ");
+        world.stdout.push(line);
+        afterLog?.(line, world);
+    },
+    error: (...args) => {
+        world.stderr.push(args.map(String).join(" "));
+    },
+    assert: () => { },
+    clear: () => { },
+    count: () => { },
+    countReset: () => { },
+    debug: () => { },
+    dir: () => { },
+    dirxml: () => { },
+    group: () => { },
+    groupCollapsed: () => { },
+    groupEnd: () => { },
+    info: () => { },
+    table: () => { },
+    time: () => { },
+    timeEnd: () => { },
+    timeLog: () => { },
+    trace: () => { },
+    warn: () => { }
+});
+/**
+ * Build a fully closed layer with the same services `HeadlessLayer` provides, plus the world the
+ * assertions read.
+ */
+export const makeFakeHeadless = (options = {}) => {
+    let jiraAccessTokenOverride;
+    let jiraAccountIdOverride;
+    let jiraCloudIdOverride;
+    let jiraSiteUrl = "https://fake.atlassian.net";
+    const world = {
+        createdClockifyEntries: [],
+        clockifyRequests: [],
+        clockifyAuth: {
+            baseUrl: "https://api.clockify.me/api",
+            workspaceId: FAKE_WORKSPACE_ID,
+            userId: FAKE_USER_ID,
+            apiKey: "key",
+            ...options.clockifyAuth
+        },
+        jiraAuth: {
+            get accessToken() {
+                return jiraAccessTokenOverride ?? options.jiraAccessToken ?? "jira-token";
+            },
+            set accessToken(value) {
+                jiraAccessTokenOverride = value;
+            },
+            get accountId() {
+                return jiraAccountIdOverride ?? options.jiraAccountId ?? FAKE_ACCOUNT_ID;
+            },
+            set accountId(value) {
+                jiraAccountIdOverride = value;
+            },
+            get cloudId() {
+                return jiraCloudIdOverride ?? options.jiraCloudId ?? "cloud-fake";
+            },
+            set cloudId(value) {
+                jiraCloudIdOverride = value;
+            },
+            get siteUrl() {
+                return jiraSiteUrl;
+            },
+            set siteUrl(value) {
+                jiraSiteUrl = value;
+            }
+        },
+        jiraRequests: [],
+        clockifyVerifiedUserId: options.clockifyVerifiedUserId ?? FAKE_USER_ID,
+        clockifyUserReadFailuresRemaining: 0,
+        clockifyEntryReadFailuresRemaining: 0,
+        clockifyEntriesReadFailuresRemaining: 0,
+        jiraWorklogs: [],
+        updatedClockifyEntries: [],
+        updatedJiraWorklogs: [],
+        attributorRequests: [],
+        attributorBatches: [],
+        describeRequests: [],
+        describeBatches: [],
+        maxAttributorInFlight: 0,
+        stdout: [],
+        stderr: [],
+        prompts: [],
+        transcripts: { ...options.transcripts },
+        transcriptReads: [],
+        // Ordinary fake scenarios model an installation whose earlier provider history was already
+        // reviewed. Pass an explicit empty map to exercise first-run migration and its fail-closed path.
+        writtenFiles: options.writtenFiles === undefined ?
+            {
+                [`${FAKE_HOME}/.jcf/source-consumption.v1.json`]: JSON.stringify({
+                    version: 3,
+                    reviewedWindows: [
+                        {
+                            provider: "clockify",
+                            scope: JSON.stringify([
+                                "clockify-v3",
+                                options.clockifyAuth?.baseUrl ?? "https://api.clockify.me/api",
+                                options.clockifyAuth?.workspaceId ?? FAKE_WORKSPACE_ID,
+                                options.clockifyVerifiedUserId ?? FAKE_USER_ID
+                            ]),
+                            fromMs: 0,
+                            toMs: 4_102_444_800_000
+                        },
+                        {
+                            provider: "jira",
+                            scope: JSON.stringify([options.jiraCloudId ?? "cloud-fake", options.jiraAccountId ?? FAKE_ACCOUNT_ID]),
+                            fromMs: 0,
+                            toMs: 4_102_444_800_000
+                        }
+                    ],
+                    pending: [],
+                    bindings: [],
+                    observedUnbound: []
+                })
+            } :
+            { ...options.writtenFiles },
+        jiraLoggedIn: options.jiraLoggedIn ?? true,
+        runningTimer: options.runningTimer ?? null,
+        jiraLoginStateReadFailuresRemaining: 0,
+        jiraCurrentUserReadFailuresRemaining: 0,
+        jiraWorklogReadFailuresRemaining: 0,
+        jiraSearchFailuresRemaining: 0,
+        jiraSearchHiddenIssues: new Set(options.jiraSearchHiddenIssues),
+        setClockifyEntries: (entries) => {
+            clockifyLedger.splice(0, clockifyLedger.length, ...entries.map((entry, index) => makeTimeEntry(entry, `replaced-${index}`)));
+        },
+        setJiraWorklogs: (entries) => {
+            jiraLedger.clear();
+            for (const [key, worklogs] of Object.entries(entries)) {
+                jiraLedger.set(key, [...worklogs]);
+                for (const entry of worklogs) {
+                    if (entry.id !== undefined)
+                        issuedJiraIds.add(entry.id);
+                }
+            }
+        }
+    };
+    // Ledgers, not fixed fixtures: a write lands here and the next read sees it, which is what
+    // makes "run it twice and nothing is proposed the second time" a real assertion.
+    const clockifyLedger = (options.clockifyEntries ?? []).map((entry, index) => makeTimeEntry(entry, `existing-${index}`));
+    const jiraLedger = new Map(Object.entries(options.jiraWorklogs ?? {}).map(([key, worklogs]) => [key, [...worklogs]]));
+    const jiraLedgersByCloudId = new Map(Object.entries(options.jiraWorklogsByCloudId ?? {}).map(([cloudId, issues]) => [
+        cloudId,
+        new Map(Object.entries(issues).map(([key, worklogs]) => [key, [...worklogs]]))
+    ]));
+    const jiraLedgerForUrl = (url) => {
+        const cloudId = url.match(/\/ex\/jira\/([^/]+)\//u)?.[1];
+        return cloudId === undefined ? jiraLedger : jiraLedgersByCloudId.get(decodeURIComponent(cloudId)) ?? jiraLedger;
+    };
+    const issuedJiraIds = new Set([jiraLedger, ...jiraLedgersByCloudId.values()].flatMap((ledger) => [...ledger.values()].flatMap((entries) => entries.flatMap((entry) => entry.id === undefined ? [] : [entry.id]))));
+    const runningEntry = () => world.runningTimer === null ? null : makeTimeEntry(world.runningTimer, "running-1");
+    const clockify = {
+        getUser: () => Effect.succeed({
+            id: FAKE_USER_ID,
+            name: "Fake",
+            email: "fake@example.com",
+            activeWorkspace: FAKE_WORKSPACE_ID,
+            defaultWorkspace: FAKE_WORKSPACE_ID,
+            profilePicture: "",
+            status: "ACTIVE"
+        }),
+        getWorkspaces: () => Effect.succeed([{ id: FAKE_WORKSPACE_ID, name: "WS", imageUrl: "" }]),
+        getProjects: () => Effect.succeed(options.clockifyProjects ?? []),
+        getProjectByName: () => Effect.succeed(null),
+        // Paged when a test asks for it, so "reads only the first page" is observable rather than
+        // assumed. `getRunningTimer` has its own path, so the running entry rides along on page one.
+        getTimeEntries: (_ws, _user, params) => {
+            if (world.clockifyEntriesReadFailuresRemaining > 0) {
+                world.clockifyEntriesReadFailuresRemaining--;
+                return Effect.fail(new HttpClientError.HttpClientError({
+                    reason: new HttpClientError.TransportError({
+                        request: HttpClientRequest.get("https://fake.clockify/time-entries"),
+                        description: "Clockify entries are unavailable"
+                    })
+                }));
+            }
+            const running = runningEntry();
+            const all = running === null ? [...clockifyLedger] : [...clockifyLedger, running];
+            const page = params?.page ?? 1;
+            const cap = options.clockifyPageSize;
+            // One page holds everything, and the page after it is empty — a paginating caller has to be
+            // able to reach the end. Returning `all` for every page would model a server that never ends.
+            if (cap === undefined)
+                return Effect.succeed(page === 1 ? all : []);
+            // The smaller of the two, because Clockify may serve fewer than it was asked for — which is
+            // the case a caller that stops at the first short page gets wrong.
+            const size = Math.min(params?.pageSize ?? cap, cap);
+            return Effect.succeed(all.slice((page - 1) * size, page * size));
+        },
+        getRunningTimer: () => options.clockifyRunningTimerReadFails === true
+            ? Effect.fail(new HttpClientError.HttpClientError({
+                reason: new HttpClientError.TransportError({
+                    request: HttpClientRequest.get("https://api.clockify.me/running-timer"),
+                    description: "Clockify running timer is unavailable"
+                })
+            }))
+            : Effect.sync(runningEntry),
+        getTimeEntry: (_ws, id) => Effect.suspend(() => {
+            if (world.clockifyEntryReadFailuresRemaining > 0) {
+                world.clockifyEntryReadFailuresRemaining--;
+                return Effect.fail(new HttpClientError.HttpClientError({
+                    reason: new HttpClientError.TransportError({
+                        request: HttpClientRequest.get(`https://fake.clockify/entries/${id}`),
+                        description: "Clockify entry read is unavailable"
+                    })
+                }));
+            }
+            const entry = clockifyLedger.find((entry) => entry.id === id);
+            if (entry !== undefined)
+                return Effect.succeed(entry);
+            const request = HttpClientRequest.get(`https://fake.clockify/entries/${id}`);
+            return Effect.fail(new HttpClientError.HttpClientError({
+                reason: new HttpClientError.StatusCodeError({
+                    request,
+                    response: HttpClientResponse.fromWeb(request, new Response(null, { status: 404 }))
+                })
+            }));
+        }),
+        getTags: () => Effect.succeed(options.clockifyTags ?? []),
+        createTag: (_ws, name) => Effect.succeed({ id: `tag-${name}`, name, workspaceId: FAKE_WORKSPACE_ID, archived: false }),
+        findOrCreateTag: (_ws, name) => Effect.succeed({ id: `tag-${name}`, name, workspaceId: FAKE_WORKSPACE_ID, archived: false }),
+        createTimeEntry: (_ws, params) => Effect.gen(function* () {
+            if (options.beforeClockifyWrite !== undefined)
+                yield* options.beforeClockifyWrite(world);
+            if (options.clockifyWritesFail === true) {
+                return yield* new HttpClientError.HttpClientError({
+                    reason: new HttpClientError.TransportError({
+                        request: HttpClientRequest.post("https://api.clockify.me/time-entries"),
+                        description: "Clockify refused the entry"
+                    })
+                });
+            }
+            return yield* Effect.sync(() => {
+                world.createdClockifyEntries.push({
+                    description: params.description,
+                    start: params.start,
+                    end: params.end,
+                    billable: params.billable
+                });
+                const entry = makeTimeEntry({ description: params.description, start: params.start, end: params.end }, `created-${clockifyLedger.length}`);
+                clockifyLedger.push(entry);
+                options.afterClockifyWrite?.(world);
+                return entry;
+            });
+        }),
+        updateTimeEntry: (_ws, id, params) => Effect.sync(() => {
+            const index = clockifyLedger.findIndex((entry) => entry.id === id);
+            const existing = clockifyLedger[index];
+            const saved = makeTimeEntry({
+                ...existing,
+                ...params,
+                description: params.description ?? existing?.description ?? "",
+                start: params.start
+            }, id);
+            world.updatedClockifyEntries.push({ id, payload: params });
+            if (index >= 0)
+                clockifyLedger[index] = saved;
+            return saved;
+        }),
+        deleteTimeEntry: () => Effect.void,
+        stopTimer: (_ws, _user, params) => Effect.succeed(makeTimeEntry({ description: "", start: "", end: params.end }, "stopped-1"))
+    };
+    // Read through the world, not captured, so one world can model logging back in between two runs —
+    // which is exactly the flow the "Jira refused, go and log in, restart" message asks the user for.
+    const loggedIn = () => world.jiraLoggedIn;
+    const jsonResponse = (request, status, body) => HttpClientResponse.fromWeb(request, new Response(JSON.stringify(body), { status, headers: { "content-type": "application/json" } }));
+    const bodyFailureResponse = (request, status) => HttpClientResponse.fromWeb(request, new Response(new ReadableStream({ start: (controller) => controller.error("synthetic body read failure") }), {
+        status,
+        headers: { "content-type": "application/json" }
+    }));
+    const requestPayload = (body) => {
+        if (body._tag !== "Uint8Array" || !("body" in body) || !Predicate.isUint8Array(body.body))
+            return {};
+        return Option.getOrElse(decodeWorklogPayload(new TextDecoder().decode(body.body)), () => ({}));
+    };
+    /**
+     * The single Jira boundary. Both the generated `JiraApiClient` (reads) and TimerService's raw
+     * worklog POST (writes) go through it, so reads exercise the real client's decoding rather than
+     * a hand-shaped stub of a very large generated interface.
+     */
+    const httpClientLayer = Layer.succeed(HttpClient.HttpClient, HttpClient.make((request) => Effect.sync(() => {
+        const requestJiraLedger = jiraLedgerForUrl(request.url);
+        if (request.url.includes("atlassian")) {
+            world.jiraRequests.push({ method: request.method, url: request.url });
+        }
+        if (request.url.includes("/v1/user") && request.method === "GET") {
+            world.clockifyRequests.push({ method: request.method, url: request.url });
+            if (options.clockifyUserReadFails === true) {
+                return jsonResponse(request, 503, { message: "synthetic verification failure" });
+            }
+            if (world.clockifyUserReadFailuresRemaining > 0) {
+                world.clockifyUserReadFailuresRemaining--;
+                return jsonResponse(request, 503, { message: "synthetic verification failure" });
+            }
+            return jsonResponse(request, 200, {
+                id: world.clockifyVerifiedUserId,
+                name: "Fake",
+                email: "fake@example.com",
+                profilePicture: "",
+                status: "ACTIVE"
+            });
+        }
+        const clockifyEntries = request.url.match(/\/v1\/workspaces\/([^/]+)\/user\/([^/]+)\/time-entries/);
+        if (clockifyEntries !== null && request.method === "GET") {
+            world.clockifyRequests.push({ method: request.method, url: request.url });
+            if (world.clockifyEntriesReadFailuresRemaining > 0) {
+                world.clockifyEntriesReadFailuresRemaining--;
+                return jsonResponse(request, 503, { message: "synthetic entry read failure" });
+            }
+            if (options.clockifyRunningTimerReadFails === true &&
+                request.urlParams.params.some(([name]) => name === "in-progress")) {
+                return jsonResponse(request, 503, { message: "synthetic timer read failure" });
+            }
+            const page = Number(request.urlParams.params.find(([name]) => name === "page")?.[1] ?? 1);
+            const cap = options.clockifyPageSize;
+            const size = Math.min(Number(request.urlParams.params.find(([name]) => name === "page-size")?.[1] ?? 200), cap ?? 200);
+            const running = runningEntry();
+            const hiddenIds = new Set(options.clockifyListHiddenIds);
+            const visibleLedger = clockifyLedger.filter((entry) => entry.id === undefined || !hiddenIds.has(entry.id));
+            const all = request.urlParams.params.some(([name]) => name === "in-progress")
+                ? running === null ? [] : [running]
+                : running === null
+                    ? visibleLedger
+                    : [...visibleLedger, running];
+            const entries = all.slice((page - 1) * size, page * size);
+            options.afterClockifyEntriesRead?.(page, world);
+            return jsonResponse(request, 200, entries);
+        }
+        const exactClockifyEntry = request.url.match(/\/v1\/workspaces\/([^/]+)\/time-entries\/([^/?]+)/);
+        if (exactClockifyEntry !== null && request.method === "GET") {
+            world.clockifyRequests.push({ method: request.method, url: request.url });
+            const fault = options.clockifyExactEntryReadFault;
+            if (fault === "malformed-404") {
+                return HttpClientResponse.fromWeb(request, new Response("{", {
+                    status: 404,
+                    headers: { "content-type": "application/json" }
+                }));
+            }
+            if (fault === "body-failure-404")
+                return bodyFailureResponse(request, 404);
+            if (fault !== undefined)
+                return jsonResponse(request, Number(fault), { message: "synthetic read failure" });
+            const entry = clockifyLedger.find((candidate) => candidate.id === exactClockifyEntry[2]);
+            return entry === undefined
+                ? jsonResponse(request, 404, { message: "synthetic time entry not found" })
+                : jsonResponse(request, 200, entry);
+        }
+        if (request.url.match(/\/v1\/workspaces\/[^/]+\/time-entries$/) !== null && request.method === "POST") {
+            world.clockifyRequests.push({ method: request.method, url: request.url });
+            if (options.clockifyWritesFail === true)
+                return jsonResponse(request, 503, { message: "synthetic failure" });
+            const body = request.body;
+            if (body._tag !== "Uint8Array" || !("body" in body) || !Predicate.isUint8Array(body.body)) {
+                return jsonResponse(request, 400, { message: "missing body" });
+            }
+            const payload = Schema.decodeUnknownOption(Schema.fromJsonString(Schema.Struct({
+                description: Schema.String,
+                start: Schema.String,
+                end: Schema.String,
+                billable: Schema.Boolean
+            })))(new TextDecoder().decode(body.body));
+            if (Option.isNone(payload))
+                return jsonResponse(request, 400, { message: "invalid body" });
+            const entry = makeTimeEntry(payload.value, `created-${clockifyLedger.length}`);
+            clockifyLedger.push(entry);
+            world.createdClockifyEntries.push(payload.value);
+            options.afterClockifyWrite?.(world);
+            return jsonResponse(request, 201, entry);
+        }
+        const exactWorklog = request.url.match(/issue\/([^/]+)\/worklog\/([^/?]+)/);
+        if (exactWorklog !== null) {
+            const issueKey = exactWorklog[1] ?? "";
+            const id = exactWorklog[2] ?? "";
+            const entries = requestJiraLedger.get(issueKey) ?? [];
+            const index = entries.findIndex((entry, index) => (entry.id ?? `wl-${index}`) === id);
+            const entry = entries[index];
+            const fault = options.jiraExactWorklogReadFault;
+            if (fault === "malformed-404") {
+                return HttpClientResponse.fromWeb(request, new Response("{", {
+                    status: 404,
+                    headers: { "content-type": "application/json" }
+                }));
+            }
+            if (fault === "body-failure-404")
+                return bodyFailureResponse(request, 404);
+            if (fault === "permission-404") {
+                return jsonResponse(request, 404, { errorMessages: ["synthetic visibility denied"] });
+            }
+            if (fault === "disabled-404") {
+                return jsonResponse(request, 404, { errorMessages: ["synthetic time tracking unavailable"] });
+            }
+            if (fault === "null-404")
+                return jsonResponse(request, 404, null);
+            if (fault === "empty-404")
+                return jsonResponse(request, 404, {});
+            if (fault !== undefined)
+                return jsonResponse(request, Number(fault), { errorMessages: ["synthetic failure"] });
+            if (entry === undefined)
+                return jsonResponse(request, 404, { errorMessages: ["Worklog not found"] });
+            if (request.method === "PUT") {
+                const payload = requestPayload(request.body);
+                entries[index] = {
+                    ...entry,
+                    started: payload.started ?? entry.started,
+                    timeSpentSeconds: payload.timeSpentSeconds ?? entry.timeSpentSeconds,
+                    ...(payload.comment !== undefined && { comment: payload.comment })
+                };
+                world.updatedJiraWorklogs.push({
+                    issueKey,
+                    id,
+                    payload,
+                    adjustEstimate: [...request.urlParams].find(([key]) => key === "adjustEstimate")?.[1]
+                });
+            }
+            return jsonResponse(request, 200, {
+                author: { accountId: options.jiraAccountId ?? FAKE_ACCOUNT_ID },
+                ...entries[index],
+                id: options.jiraReadOmitsId === true ? undefined : options.jiraExactWorklogReturnedId ?? id
+            });
+        }
+        const worklogMatch = request.url.match(/issue\/([^/]+)\/worklog/);
+        if (request.method === "POST" && worklogMatch !== null) {
+            const payload = requestPayload(request.body);
+            const issueKey = worklogMatch[1] ?? "unknown";
+            let nextId = 0;
+            while (issuedJiraIds.has(`wl-created-${nextId}`))
+                nextId++;
+            const entryId = options.jiraPostedWorklog?.id ?? `wl-created-${nextId}`;
+            if (issuedJiraIds.has(entryId)) {
+                return jsonResponse(request, 409, { errorMessages: ["Synthetic worklog ID already exists"] });
+            }
+            issuedJiraIds.add(entryId);
+            const started = payload.started ?? "";
+            const timeSpentSeconds = payload.timeSpentSeconds ?? 0;
+            world.jiraWorklogs.push({ issueKey, started, timeSpentSeconds, comment: commentText(payload.comment) });
+            requestJiraLedger.set(issueKey, [...(requestJiraLedger.get(issueKey) ?? []), {
+                    id: entryId,
+                    ...(options.jiraPostedWorklog !== undefined && {
+                        author: { accountId: options.jiraPostedWorklog.accountId },
+                        ...(options.jiraPostedWorklog.created !== undefined && { created: options.jiraPostedWorklog.created })
+                    }),
+                    started,
+                    timeSpentSeconds,
+                    ...(payload.comment !== undefined && { comment: payload.comment })
+                }]);
+            return jsonResponse(request, 201, options.jiraPostOmitsId === true ? {} : { id: entryId });
+        }
+        if (worklogMatch !== null) {
+            const requestedStartAt = Number(request.urlParams.params.find(([name]) => name === "startAt")?.[1] ?? 0);
+            if (requestedStartAt > 0 && options.jiraWorklogPageFault === "failure") {
+                return jsonResponse(request, 500, { errorMessages: ["Later worklog page unavailable"] });
+            }
+            if (options.jiraWorklogReadFails === true || world.jiraWorklogReadFailuresRemaining > 0) {
+                if (world.jiraWorklogReadFailuresRemaining > 0)
+                    world.jiraWorklogReadFailuresRemaining--;
+                return jsonResponse(request, 500, { errorMessages: ["Jira is having a moment"] });
+            }
+            const numberParam = (name) => {
+                const raw = request.urlParams.params.find(([candidate]) => candidate === name)?.[1];
+                if (raw === undefined)
+                    return undefined;
+                const parsed = Number(raw);
+                return Number.isFinite(parsed) ? parsed : undefined;
+            };
+            const startedAfter = numberParam("startedAfter");
+            const startedBefore = numberParam("startedBefore");
+            const startAt = numberParam("startAt") ?? 0;
+            const entries = (requestJiraLedger.get(worklogMatch[1] ?? "") ?? []).filter((worklog) => {
+                const startedMs = new Date(worklog.started).getTime();
+                return (startedAfter === undefined || startedMs >= startedAfter) &&
+                    (startedBefore === undefined || startedMs < startedBefore);
+            });
+            const pageSize = options.jiraWorklogPageSize ?? entries.length;
+            const page = entries.slice(startAt, startAt + pageSize);
+            if (startAt > 0 && options.jiraWorklogPageFault === "missing") {
+                return jsonResponse(request, 200, { startAt, maxResults: pageSize, total: entries.length });
+            }
+            const response = {
+                startAt: startAt > 0 && options.jiraWorklogPageFault === "stalled" ? 0 : startAt,
+                maxResults: pageSize,
+                total: entries.length + (startAt > 0 && options.jiraWorklogPageFault === "inconsistent" ? 1 : 0),
+                worklogs: page.map((worklog, index) => {
+                    const id = worklog.id ?? `wl-${startAt + index}`;
+                    issuedJiraIds.add(id);
+                    return {
+                        id,
+                        author: { accountId: options.jiraAccountId ?? FAKE_ACCOUNT_ID },
+                        ...worklog,
+                        ...(startAt > 0 && options.jiraWorklogPageFault === "missing-duration" && {
+                            timeSpentSeconds: undefined
+                        }),
+                        ...(startAt > 0 && options.jiraWorklogPageFault === "missing-started" && { started: undefined }),
+                        ...(startAt > 0 && options.jiraWorklogPageFault === "missing-author" && { author: undefined }),
+                        ...(startAt > 0 && options.jiraWorklogPageFault === "missing-id" && { id: undefined }),
+                        ...(startAt > 0 && options.jiraWorklogPageFault === "empty-id" && { id: "" }),
+                        ...(startAt > 0 && options.jiraWorklogPageFault === "spaces-id" && { id: "   " }),
+                        ...(startAt > 0 && options.jiraWorklogPageFault === "tab-newline-id" && { id: "\t\n" }),
+                        ...(options.jiraReadOmitsId === true && { id: undefined })
+                    };
+                })
+            };
+            options.afterJiraWorklogsRead?.(worklogMatch[1] ?? "", startAt, world);
+            return jsonResponse(request, 200, response);
+        }
+        const issueMatch = request.url.match(/issue\/([^/?]+)(?:\?|$)/);
+        if (request.method === "GET" && issueMatch !== null) {
+            const key = issueMatch[1] ?? "";
+            const summary = (options.issueSummaries ?? {})[key];
+            if (summary === undefined) {
+                return jsonResponse(request, 404, { errorMessages: ["Issue does not exist"] });
+            }
+            const assignee = (options.issueAssignees ?? {})[key];
+            const displayName = (options.issueAssigneeDisplayNames ?? {})[key];
+            return jsonResponse(request, 200, {
+                id: "10000",
+                key,
+                fields: {
+                    summary,
+                    ...((assignee !== undefined) && {
+                        assignee: assignee === null ? null : displayName === null
+                            ? {}
+                            : { displayName: displayName ?? assignee }
+                    })
+                }
+            });
+        }
+        if (request.url.includes("/myself")) {
+            if (options.jiraCurrentUserFails === true || world.jiraCurrentUserReadFailuresRemaining > 0) {
+                if (world.jiraCurrentUserReadFailuresRemaining > 0)
+                    world.jiraCurrentUserReadFailuresRemaining--;
+                return jsonResponse(request, 500, { errorMessages: ["Current user unavailable"] });
+            }
+            return jsonResponse(request, 200, {
+                accountId: world.jiraAuth.accountId,
+                displayName: "Fake User"
+            });
+        }
+        if (request.url.includes("/search/jql")) {
+            if (world.jiraSearchFailuresRemaining > 0) {
+                world.jiraSearchFailuresRemaining--;
+                return jsonResponse(request, 500, { errorMessages: ["Jira search is unavailable"] });
+            }
+            // A `key in (…)` search answers about exactly those keys. Any other JQL keeps the old
+            // behaviour: whatever the ledger happens to hold. The query lives in `urlParams` rather
+            // than in `url` — the client keeps them apart until the request is actually sent.
+            const jql = request.urlParams.params.find(([name]) => name === "jql")?.[1];
+            const clause = jql === undefined ? null : jql.match(/key in \(([^)]*)\)/);
+            if (clause !== null) {
+                const keys = (clause[1] ?? "").split(",").map((key) => key.trim()).filter((key) => key !== "");
+                const summaries = options.issueSummaries ?? {};
+                const assignees = options.issueAssignees ?? {};
+                const displayNames = options.issueAssigneeDisplayNames ?? {};
+                return jsonResponse(request, 200, {
+                    // Only the keys this fake Jira knows about, so a test can model an issue that is gone.
+                    issues: keys.filter((key) => summaries[key] !== undefined).map((key, index) => ({
+                        id: String(index),
+                        key,
+                        fields: {
+                            summary: summaries[key],
+                            ...((assignees[key] !== undefined) && {
+                                assignee: assignees[key] === null
+                                    ? null
+                                    : displayNames[key] === null
+                                        ? { accountId: assignees[key] }
+                                        : { accountId: assignees[key], displayName: displayNames[key] ?? assignees[key] }
+                            })
+                        }
+                    }))
+                });
+            }
+            return jsonResponse(request, 200, {
+                ...(options.jiraSearchClaimsMoreWithoutToken === true && { isLast: false }),
+                issues: [...requestJiraLedger.keys()]
+                    .filter((key) => !world.jiraSearchHiddenIssues.has(key))
+                    .map((key, index) => ({ id: String(index), key }))
+            });
+        }
+        return jsonResponse(request, 404, { errorMessages: [`unrouted: ${request.method} ${request.url}`] });
+    })));
+    let attributorInFlight = 0;
+    const attributorLayer = Layer.succeed(SessionAttributor, {
+        attribute: (requests) => Effect.gen(function* () {
+            for (const request of requests) {
+                world.attributorRequests.push({ candidateKeys: request.candidateKeys, digest: request.digest });
+            }
+            world.attributorBatches.push(requests.map((request) => request.sessionId));
+            attributorInFlight += 1;
+            world.maxAttributorInFlight = Math.max(world.maxAttributorInFlight, attributorInFlight);
+            // Yield so sibling calls get a chance to start. A real call is a CLI process taking tens of
+            // seconds, so overlapping is the whole point; without a suspension point here every fake
+            // call would complete before the next began and the high-water mark could not tell a
+            // concurrent implementation from a sequential one.
+            //
+            // Cooperative yielding, not `Effect.sleep`: `it.effect` runs on a TestClock that nothing
+            // here advances, so a sleep would never return.
+            yield* Effect.yieldNow;
+            yield* Effect.yieldNow;
+            attributorInFlight -= 1;
+            const answers = [];
+            for (const request of requests) {
+                const answer = options.attributor?.(request) ?? { _tag: "None" };
+                // A single "fail" fails the whole call, which is what a real timeout does to a batch.
+                if (answer === "fail") {
+                    return yield* new SessionAttributorError({ message: "claude executable not found" });
+                }
+                answers.push({ sessionId: request.sessionId, choice: answer });
+            }
+            return answers;
+        }),
+        describe: (requests) => Effect.gen(function* () {
+            options.beforeDescribe?.(world);
+            for (const request of requests) {
+                world.describeRequests.push({
+                    ticketKey: request.ticketKey,
+                    summary: request.summary,
+                    digest: request.digest
+                });
+            }
+            world.describeBatches.push(requests.map((request) => request.ticketKey));
+            const answers = [];
+            for (const request of requests) {
+                const answer = options.describer?.(request) ?? null;
+                // As with attribution, one "fail" fails the whole call — a timeout costs a batch.
+                if (answer === "fail") {
+                    return yield* new SessionAttributorError({ message: "claude executable not found" });
+                }
+                answers.push({ id: request.id, note: answer });
+            }
+            return answers;
+        })
+    });
+    /**
+     * A spawner that refuses to spawn.
+     *
+     * The Coding Agent is faked at the `SessionAttributor` seam, so nothing in these tests should
+     * reach a child process — but the layer was previously left open, which typechecking would have
+     * caught had this package's tests been in a tsconfig. Failing loudly beats a real `claude` process
+     * starting during a test run.
+     */
+    const SpawnerLayer = Layer.succeed(ChildProcessSpawner.ChildProcessSpawner, ChildProcessSpawner.make(() => Effect.die("the fake headless layer spawns no child processes")));
+    const HomeLayer = Layer.succeed(HomeDirectory, { path: FAKE_HOME });
+    // A ledger rather than a constant: `set` is how the config commands do their whole job, so a
+    // fake that discards it can only ever assert what was printed, not what was stored.
+    let config = { ...defaultConfig, ...options.config };
+    const ConfigLayer = Layer.succeed(ConfigService, {
+        get: Effect.gen(function* () {
+            const snapshot = config;
+            if (options.delayConfigReads === true)
+                yield* Effect.sleep("10 millis");
+            return snapshot;
+        }),
+        set: (patch) => Effect.sync(() => {
+            config = { ...config, ...patch };
+        }),
+        configDir: Effect.succeed(`${FAKE_HOME}/.jcf`)
+    });
+    const ClockifyAuthLayer = Layer.succeed(ClockifyAuth, {
+        getConfig: Effect.sync(() => ({
+            ...world.clockifyAuth,
+            apiKey: Redacted.make(world.clockifyAuth.apiKey)
+        })),
+        save: () => Effect.void,
+        isConfigured: Effect.succeed(true)
+    });
+    const JiraAuthLayer = Layer.succeed(JiraAuth, {
+        configure: () => Effect.void,
+        isConfigured: () => Effect.succeed(loggedIn()),
+        login: () => Effect.void,
+        logout: () => Effect.void,
+        getAccessToken: () => Effect.succeed(Redacted.make(loggedIn() ? world.jiraAuth.accessToken : "")),
+        getCloudId: () => Effect.succeed(world.jiraAuth.cloudId),
+        getSiteUrl: () => Effect.succeed(world.jiraAuth.siteUrl),
+        getCurrentUser: () => Effect.succeed(!loggedIn() || options.jiraCachedUserMissing === true ? null : {
+            account_id: world.jiraAuth.accountId,
+            name: "Fake User",
+            email: "fake@example.com"
+        }),
+        getActiveProfile: () => Effect.succeed(loggedIn() ?
+            {
+                id: "synthetic-profile",
+                name: "Synthetic",
+                token: {
+                    access_token: world.jiraAuth.accessToken,
+                    refresh_token: "synthetic-refresh",
+                    expires_at: 4_102_444_800_000,
+                    scope: "write:jira-work",
+                    cloud_id: world.jiraAuth.cloudId,
+                    site_url: world.jiraAuth.siteUrl,
+                    user: {
+                        account_id: world.jiraAuth.accountId,
+                        name: "Fake User",
+                        email: "fake@example.com"
+                    }
+                },
+                created_at: "2026-01-01T00:00:00.000Z",
+                updated_at: "2026-01-01T00:00:00.000Z"
+            } :
+            null),
+        listProfiles: () => Effect.succeed([]),
+        switchProfile: () => Effect.succeed(null),
+        removeProfile: () => Effect.succeed(null),
+        isLoggedIn: () => {
+            if (world.jiraLoginStateReadFailuresRemaining === 0)
+                return Effect.succeed(loggedIn());
+            world.jiraLoginStateReadFailuresRemaining--;
+            return Effect.fail(systemError({
+                _tag: "PermissionDenied",
+                module: "FileSystem",
+                method: "readFileString",
+                description: "synthetic Jira login state read failure",
+                pathOrDescriptor: `${FAKE_HOME}/.jira/config.json`
+            }));
+        }
+    });
+    const StateWriterLayer = Layer.succeed(StateWriter, {
+        write: () => Effect.void,
+        read: Effect.succeed({
+            active: false,
+            ticketKey: null,
+            summary: null,
+            project: null,
+            startedAt: null,
+            startedAt_unix: null,
+            elapsed: 0,
+            clockifyEntryId: null
+        }),
+        clear: Effect.void
+    });
+    const ClockifyLayer = Layer.succeed(ClockifyApiClient, clockify);
+    const resolvedJiraCredential = () => ({
+        type: "oauth2",
+        accessToken: Redacted.make(world.jiraAuth.accessToken),
+        cloudId: world.jiraAuth.cloudId
+    });
+    const jiraConfig = options.beforeJiraAuthResolution === undefined
+        ? {
+            baseUrl: "https://fake.atlassian.net",
+            auth: { type: "basic", email: "fake@example.com", apiToken: Redacted.make("token") }
+        }
+        : {
+            baseUrl: "",
+            auth: resolvedJiraCredential(),
+            resolveAuth: Effect.sync(() => {
+                options.beforeJiraAuthResolution?.(world);
+                return resolvedJiraCredential();
+            })
+        };
+    const JiraConfigLayer = Layer.succeed(JiraApiConfig, jiraConfig);
+    const FileSystemLayer = fakeFileSystemLayer(world.transcripts, world.transcriptReads, world.writtenFiles, options.unwritablePaths ?? [], options.unreadableTranscripts ?? [], options.afterFileWrite ?? (() => Effect.void), options.pidNamespaceInode);
+    const Externals = Layer.mergeAll(ClockifyLayer, ClockifyAuthLayer, JiraAuthLayer, ConfigLayer, StateWriterLayer, HomeLayer, FileSystemLayer, NodePath.layer, 
+    // The CLI runner writes help and error text through Stdio; the command's own output goes
+    // through the captured Console instead.
+    Stdio.layerTest({}), httpClientLayer, attributorLayer, fakeTerminalLayer({
+        keep: options.keep,
+        promptInputs: options.promptInputs,
+        columns: options.columns ?? DEFAULT_COLUMNS,
+        beforePromptInput: options.beforePromptInput
+    }, world), captureConsoleLayer(world, options.afterConsoleLog), SpawnerLayer, 
+    // Mirrors HeadlessLayer: warnings must land on stderr, or they corrupt --json output.
+    LogToStderrLive);
+    // The real generated client over the routed fake HttpClient, so read decoding is exercised.
+    const JiraLayer = JiraApiClient.layer.pipe(Layer.provide(JiraConfigLayer), Layer.provide(httpClientLayer));
+    const TimerLive = timerServiceLayer.pipe(Layer.provide(Externals));
+    const SourceLedgerLive = sourceLedgerLayer.pipe(Layer.provide(Externals));
+    const ReaderLive = agentSessionReaderLayer.pipe(Layer.provide(Externals));
+    const ReconcileLive = reconcileServiceLayer.pipe(Layer.provide(Externals), Layer.provide(JiraLayer), Layer.provide(TimerLive), Layer.provide(ReaderLive), Layer.provide(SourceLedgerLive));
+    const SavedLive = savedEntriesLayer.pipe(Layer.provide(Externals), Layer.provide(JiraLayer));
+    const TicketLive = ticketServiceLayer.pipe(Layer.provide(Externals), Layer.provide(JiraLayer));
+    const IssueFactsLive = issueFactsLayer.pipe(Layer.provide(Externals), Layer.provide(JiraLayer));
+    // `ReconcileService` is built from the timer, the reader and the Jira client, so those cannot sit
+    // beside it in a `mergeAll` — that builds its members in parallel. They go underneath.
+    const layer = Layer.mergeAll(ReconcileLive, TicketLive, IssueFactsLive, SavedLive).pipe(Layer.provideMerge(Layer.mergeAll(TimerLive, ReaderLive, JiraLayer, SourceLedgerLive)), Layer.provideMerge(Externals));
+    return { layer, sourceLedgerLayer: SourceLedgerLive, world };
+};
+//# sourceMappingURL=fakeHeadless.js.map