@gitroll/tacitt-cli 0.0.0-stage → 0.2.0

package/README.md

@@ -1,3 +1,85 @@
-# Temporary Holding Version
+# SessPilot
 
-This version is a temporary placeholder for this package. An operational version to replace this has been submitted for review and is awaiting a staged release.
+Local session review and export through the public npm package. Requires Node.js
+`>=24.19.0 <25`, npm and connectivity to the npm registry. Consumers do not need npm
+login or organization membership.
+
+This release integrates B privacy analysis with the native Ray review journey and
+configured v5 export, published on both `latest` and `next`. Before any reading, the
+agent groups sessions into Works by repository (linked worktrees fold into their
+repository) named after the project, proposes a project type for each Work and lets
+you leave whole types out; Dropped Works are not read. Older cached credentials
+require a fresh Identity sign-in.
+
+This release uses storage schema 20, Journey format 8 and review contract 5. State
+from older previews, including 0.2.0, rc.1 and the b candidates, is preserved and is not
+automatically migrated or imported; use a fresh, separate state directory for a
+new review and retain the pinned CLI for older tasks. Installing or updating the
+Skill does not reset task state. Resetting control state does not erase private
+review payload files: their bodies remain locally readable, and a fresh store
+does not discover or import them.
+
+Windows verification remains partly advisory: some automated account and lifecycle
+checks still time out. Prefer macOS or Linux.
+
+Start the Manual static browser journey:
+
+```sh
+npx --yes @gitroll/tacitt-cli@latest
+```
+
+Install or deliberately update both Codex and Claude user Skills:
+
+```sh
+npx --yes @gitroll/tacitt-cli@latest skill.install
+```
+
+Add `--agent codex` or `--agent claude` to select a client. Add `--scope project`
+from a workspace to install there instead. After installation, the terminal shows
+the next step: enter `$tacitt` in a fresh Codex conversation or `/tacitt`
+in Claude Code. The Skill prepares its exact matching CLI
+through npx; no global SessPilot installation is necessary. Existing loaded tasks
+keep that release; a later deliberate install affects new invocations. Retain the
+recorded exact invocation when resuming a previous task after an agent restart.
+
+The human hint uses stderr and only lists the installed clients. Stdout remains
+JSON; explicit `--input` keeps the machine interface without a hint. Installation
+itself starts no scan.
+
+The agent handles the conversation-led journey. One Allow in chat authorizes B review through the current App's model provider. The
+provider receives locally cleaned original text **before privacy masking**:
+complete main user and assistant text and retained tool results. Tool-call
+arguments, children, setup files and other source metadata receive static
+processing; images are not analyzed. Reading starts only after consent. Every decision is made in the
+conversation; the right panel is a read-only view of progress, Works, actual output
+previews and Send scope. Only official sign-in opens a browser page.
+
+Each direct category answer applies once. Certain/protected content stays hidden;
+delegated choices are attributed to the agent for the current run. Complete reading,
+verified ZIP preview and Work review precede Send; a ready subset requires explicit scope
+confirmation. Send uses official Identity, uploads the confirmed primary and
+follows its actual assessment into the owner's current Passport. Local export
+remains available before upload. The standalone Manual journey uses its existing
+static browser interface.
+
+You can limit a category answer to the current review. It remains your own choice
+without changing preferences saved for later runs. An established current-only
+limit survives later answers; ask explicitly to remember it for future runs.
+
+The CLI and installed Skill pin the same exact version for the entire journey.
+Local ZIP success does not establish deployment compatibility or a completed
+assessment. Verify the target service's v5 support before using Send. Codex and
+Claude native App acceptance are separate checks; success in one does not verify
+the other. Local-only requests stop before Identity sign-in or upload.
+
+Updates only replace verified managed Skill files and preserve prior release
+references. Unmanaged or edited Skill collisions are reported for explicit review;
+state, dictionaries, source logs and existing ZIPs are not reset by installation.
+Package acquisition, network or version-verification failures stop setup before
+dependent session work.
+
+Bundled Web dependency notices are included in
+`dist/web/THIRD_PARTY_LICENSES.md`. SessPilot's own package license metadata remains
+`UNLICENSED`.
+
+Source, requirements and release evidence: [gitroll-dev/tacitt-cli](https://github.com/gitroll-dev/tacitt-cli).

package/bin/INDEX.md

@@ -0,0 +1,9 @@
+# Executable prerequisite index
+
+> Audience: contributors and reviewers
+> Ownership: root-maintained distribution boundary
+
+| Entry                    | Kind            | Responsibility                                                                                     |
+| ------------------------ | --------------- | -------------------------------------------------------------------------------------------------- |
+| [INDEX.md](INDEX.md)     | Directory index | Navigates the minimal executable boundary.                                                         |
+| [tacitt.cjs](tacitt.cjs) | Executable shim | Checks supported Node before loading the packaged ESM CLI and emits safe startup failure guidance. |

package/bin/tacitt.cjs

@@ -0,0 +1,16 @@
+#!/usr/bin/env node
+// This boundary must parse before loading Node24-only application modules.
+const [major, minor] = process.versions.node.split('.').map(Number)
+if (major !== 24 || minor < 19) {
+  process.stderr.write(
+    'SessPilot requires Node.js >=24.19.0 <25. Install a supported Node24 release and retry.\n',
+  )
+  process.exitCode = 1
+} else {
+  import('../dist/cli.mjs').catch(() => {
+    process.stderr.write(
+      'SESS_PILOT_STARTUP_FAILED. Reinstall the selected package release and retry.\n',
+    )
+    process.exitCode = 1
+  })
+}

package/dist/agent-setup-B8DAiQ6Z.mjs

@@ -0,0 +1,71 @@
+import "./locate-U3-nkird.mjs";
+import "./src-CC95LYYp.mjs";
+import { z } from "zod";
+import { Worker } from "node:worker_threads";
+//#region apps/cli/src/agent-setup/index.ts
+const listingSchema = z.object({
+	files: z.array(z.object({
+		key: z.string().regex(/^[a-f0-9]{64}$/u),
+		category: z.enum([
+			"project",
+			"local",
+			"user",
+			"memory"
+		]),
+		path: z.string().nullable(),
+		text: z.string(),
+		truncated: z.boolean(),
+		withheld: z.boolean()
+	})).max(256).readonly(),
+	incomplete: z.boolean()
+});
+const messageSchema = z.discriminatedUnion("kind", [z.object({
+	kind: z.literal("result"),
+	listing: listingSchema
+}), z.object({
+	kind: z.literal("error"),
+	code: z.enum(["PROCESSING_PROTECTED_CUSTOM_TERM", "PREVIEW_WITHHELD"])
+})]);
+/** The worker stops scanning at 20 seconds; this ceiling also covers startup and redaction. */
+const requestMilliseconds = 3e4;
+/** Lists agent setup files in a fresh worker so source scans never block the host loop. */
+function createAgentSetupPort(workerUrl, environment) {
+	return { list: async (request, signal) => {
+		if (signal.aborted) throw new Error("PREVIEW_CANCELED");
+		const worker = new Worker(workerUrl, {
+			workerData: {
+				request,
+				environment
+			},
+			stdout: true,
+			stderr: true,
+			resourceLimits: {
+				maxOldGenerationSizeMb: 256,
+				maxYoungGenerationSizeMb: 32
+			}
+		});
+		worker.stdout.resume();
+		worker.stderr.resume();
+		const done = Promise.withResolvers();
+		const abort = () => done.reject(/* @__PURE__ */ new Error("PREVIEW_CANCELED"));
+		const timer = setTimeout(() => done.reject(/* @__PURE__ */ new Error("PREVIEW_TIMEOUT")), requestMilliseconds);
+		signal.addEventListener("abort", abort, { once: true });
+		worker.on("message", (message) => {
+			const parsed = messageSchema.safeParse(message);
+			if (!parsed.success) done.reject(/* @__PURE__ */ new Error("PREVIEW_WITHHELD"));
+			else if (parsed.data.kind === "error") done.reject(new Error(parsed.data.code));
+			else done.resolve(parsed.data.listing);
+		});
+		worker.on("error", () => done.reject(/* @__PURE__ */ new Error("PREVIEW_WITHHELD")));
+		worker.on("exit", () => done.reject(/* @__PURE__ */ new Error("PREVIEW_WITHHELD")));
+		try {
+			return await done.promise;
+		} finally {
+			clearTimeout(timer);
+			signal.removeEventListener("abort", abort);
+			await worker.terminate();
+		}
+	} };
+}
+//#endregion
+export { createAgentSetupPort as t };

package/dist/agent-setup-worker.mjs

@@ -0,0 +1,56 @@
+import { q as listAgentSetup } from "./src-k0Ya1Bri.mjs";
+import { rn as workspaceSchema, tn as sessionSchema, yn as createStaticProcessor } from "./locate-U3-nkird.mjs";
+import { l as createNativeAdapters, u as createGitNameSeedProvider } from "./src-B5SArpVY.mjs";
+import { ct as maximumAgentSetupSessions, ni as effectiveTreatmentSchema } from "./src-CC95LYYp.mjs";
+import "./agent-setup-B8DAiQ6Z.mjs";
+import { z } from "zod";
+import { parentPort, workerData } from "node:worker_threads";
+//#region apps/cli/src/agent-setup/worker.ts
+const inputSchema = z.object({
+	request: z.object({
+		workspace: workspaceSchema,
+		sessions: z.array(z.object({
+			session: sessionSchema,
+			treatmentId: z.string().min(1).max(256)
+		})).max(maximumAgentSetupSessions),
+		treatments: z.array(effectiveTreatmentSchema).max(1e4),
+		customTerms: z.object({
+			revision: z.string().min(1).max(64),
+			terms: z.array(z.string().max(1024)).max(1e5)
+		}).optional()
+	}),
+	environment: z.object({
+		platform: z.enum(["posix", "windows"]),
+		userHome: z.string().optional(),
+		temporaryRoots: z.array(z.string())
+	})
+});
+function notify(message) {
+	parentPort?.postMessage(message);
+}
+try {
+	const data = workerData;
+	const input = inputSchema.parse(data);
+	const processing = {
+		adapters: createNativeAdapters(),
+		gitNames: createGitNameSeedProvider(),
+		processor: createStaticProcessor(),
+		environment: {
+			platform: input.environment.platform,
+			...input.environment.userHome === void 0 ? {} : { userHome: input.environment.userHome },
+			temporaryRoots: input.environment.temporaryRoots
+		},
+		limits: { maxUnitMs: 2e3 }
+	};
+	notify({
+		kind: "result",
+		listing: await listAgentSetup(input.request, processing, new AbortController().signal)
+	});
+} catch (error) {
+	notify({
+		kind: "error",
+		code: error instanceof Error && error.message === "PROCESSING_PROTECTED_CUSTOM_TERM" ? "PROCESSING_PROTECTED_CUSTOM_TERM" : "PREVIEW_WITHHELD"
+	});
+}
+//#endregion
+export {};