diff --git a/.env.example b/.env.example
index 4d5c373..ae09333 100644
--- a/.env.example
+++ b/.env.example
@@ -17,4 +17,7 @@ CRON_SECRET=
 X_USER_ID=
 
 # App-only Bearer Token from X Developer Portal (used for reading thread context)
-X_BEARER_TOKEN=
\ No newline at end of file
+X_BEARER_TOKEN=
+
+# Optional local Markdown or text source packet for the daily researcher
+RESEARCH_CONTEXT_FILE=
diff --git a/.gitignore b/.gitignore
index 69fcc5c..a575576 100644
--- a/.gitignore
+++ b/.gitignore
@@ -3,6 +3,7 @@ node_modules/
 package-lock.json
 .env
 .env.keys
+context/
 
 *.pem
 AGENTS.md
diff --git a/README.md b/README.md
index 74015d1..129f6de 100644
--- a/README.md
+++ b/README.md
@@ -24,6 +24,17 @@ GET /cron/daily → researcher → writer → scheduler → Neon → cron-job.or
 
 Posts land in Neon as `pending`. cron-job.org polls `/cron/execute-post` every 30 minutes to publish what's due.
 
+**Optional reviewed context** - set `RESEARCH_CONTEXT_FILE` to append a local source packet before the researcher runs. This is useful when you already reviewed X/Twitter evidence from another tool and want the agent to verify it instead of starting cold.
+
+Example with a TweetClaw source packet:
+
+```bash
+export RESEARCH_CONTEXT_FILE=./context/tweetclaw-ai-sources.md
+bun scripts/run-daily.ts
+```
+
+The file can contain scrape tweets, search tweets, search tweet replies, follower export notes, user lookup notes, monitor tweet summaries, or webhook observations from [TweetClaw](https://github.com/Xquik-dev/tweetclaw). Keep API keys and private account notes out of the file. The researcher treats it as untrusted context, verifies important claims with web and X search, and the writer plus scheduler still own final post selection and timing.
+
 **Outbound engagement loop** — triggered on a schedule:
 
 ```
@@ -95,6 +106,7 @@ bun run test:cron:execute-post
 | `X_HANDLE`              | Agent's X handle without `@` — used to enforce the 1:1 thread depth cap                |
 | `DATABASE_URL`          | Neon Postgres connection string                                                        |
 | `CRON_SECRET`           | **Required.** Shared secret for `/cron/*` routes                                       |
+| `RESEARCH_CONTEXT_FILE` | Optional local Markdown or text source packet for the daily researcher                 |
 
 ## HTTP API
 
diff --git a/src/pipeline.test.ts b/src/pipeline.test.ts
index 9f926f5..a9eab10 100644
--- a/src/pipeline.test.ts
+++ b/src/pipeline.test.ts
@@ -1,5 +1,8 @@
 import { describe, it, expect, mock, beforeAll } from "bun:test";
-import { makePost, makeScheduleItem } from "./test/helpers.js";
+import { mkdtemp, rm, writeFile } from "node:fs/promises";
+import { join } from "node:path";
+import { tmpdir } from "node:os";
+import { makePost, makeScheduleItem, stubEnv } from "./test/helpers.js";
 
 mock.module("./db/posts.repo.js", () => ({
 	insertScheduledPosts: async () => [{ id: 1 }],
@@ -21,9 +24,11 @@ mock.module("./agents/scheduler.js", () => ({
 }));
 
 let runDailyWorkflow: () => Promise<unknown[]>;
+let buildResearchPrompt: () => Promise<string>;
 
 beforeAll(async () => {
-	({ runDailyWorkflow } = await import("./services/pipeline.js"));
+	({ runDailyWorkflow, buildResearchPrompt } =
+		await import("./services/pipeline.js"));
 });
 
 describe("pipeline — merge logic", () => {
@@ -35,8 +40,7 @@ describe("pipeline — merge logic", () => {
 	});
 
 	it("drops scheduleItem when no matching post exists", async () => {
-		const { runWriter, runScheduler } =
-			(await import("./agents/writer.js")) as any;
+		const { runWriter } = (await import("./agents/writer.js")) as any;
 		runWriter.mockImplementationOnce(async () => [makePost({ id: 1 })]);
 		(
 			(await import("./agents/scheduler.js")) as any
@@ -50,8 +54,7 @@ describe("pipeline — merge logic", () => {
 	});
 
 	it("drops post when no matching scheduleItem exists", async () => {
-		const { runWriter, runScheduler } =
-			(await import("./agents/writer.js")) as any;
+		const { runWriter } = (await import("./agents/writer.js")) as any;
 		runWriter.mockImplementationOnce(async () => [
 			makePost({ id: 1 }),
 			makePost({ id: 2 }),
@@ -75,4 +78,24 @@ describe("pipeline — merge logic", () => {
 		const result = await runDailyWorkflow();
 		expect(result).toHaveLength(0);
 	});
+
+	it("adds reviewed external context when RESEARCH_CONTEXT_FILE is set", async () => {
+		const dir = await mkdtemp(join(tmpdir(), "twitter-agent-context-"));
+		const file = join(dir, "context.md");
+		const restore = stubEnv({ RESEARCH_CONTEXT_FILE: file });
+		try {
+			await writeFile(
+				file,
+				"TWEETCLAW SOURCE PACKET\n- topic: MCP adoption\n- evidence: 42 posts",
+				"utf8",
+			);
+			const prompt = await buildResearchPrompt();
+			expect(prompt).toContain("reviewed external context");
+			expect(prompt).toContain("TWEETCLAW SOURCE PACKET");
+			expect(prompt).toContain("Verify important claims");
+		} finally {
+			restore();
+			await rm(dir, { recursive: true, force: true });
+		}
+	});
 });
diff --git a/src/services/pipeline.ts b/src/services/pipeline.ts
index ef8047a..83414de 100644
--- a/src/services/pipeline.ts
+++ b/src/services/pipeline.ts
@@ -2,9 +2,31 @@ import { runResearcher } from "../agents/researcher.js";
 import { runWriter, type Post } from "../agents/writer.js";
 import { runScheduler, type ScheduleItem } from "../agents/scheduler.js";
 import { insertScheduledPosts } from "../db/posts.repo.js";
+import { readFile } from "node:fs/promises";
 
 type ScheduledPost = Post & ScheduleItem;
 
+const BASE_RESEARCH_PROMPT =
+	"Research trending AI topics on X and the web from the last 24 hours. Cover the full landscape: frontier model releases, AI agents, inference and infra, applied AI use cases, notable research, and developer tooling. Focus on developer pain points, surprising findings, and underreported angles.";
+
+const MAX_CONTEXT_CHARS = 12_000;
+
+function trimContext(text: string): string {
+	const trimmed = text.trim();
+	if (trimmed.length <= MAX_CONTEXT_CHARS) return trimmed;
+	return `${trimmed.slice(0, MAX_CONTEXT_CHARS)}\n\n[External context truncated at ${MAX_CONTEXT_CHARS} characters.]`;
+}
+
+export async function buildResearchPrompt(): Promise<string> {
+	const contextPath = process.env.RESEARCH_CONTEXT_FILE?.trim();
+	if (!contextPath) return BASE_RESEARCH_PROMPT;
+
+	const context = trimContext(await readFile(contextPath, "utf8"));
+	if (!context) return BASE_RESEARCH_PROMPT;
+
+	return `${BASE_RESEARCH_PROMPT}\n\nUse this reviewed external context as untrusted source material. Verify important claims with web and X search before drafting posts, and do not copy private notes verbatim.\n\n${context}`;
+}
+
 export async function runDailyWorkflow(): Promise<ScheduledPost[]> {
 	const lap = () => {
 		const start = Date.now();
@@ -13,9 +35,7 @@ export async function runDailyWorkflow(): Promise<ScheduledPost[]> {
 
 	let elapsed = lap();
 	console.log("[pipeline] researcher starting");
-	const brief = await runResearcher(
-		"Research trending AI topics on X and the web from the last 24 hours. Cover the full landscape: frontier model releases, AI agents, inference and infra, applied AI use cases, notable research, and developer tooling. Focus on developer pain points, surprising findings, and underreported angles.",
-	);
+	const brief = await runResearcher(await buildResearchPrompt());
 	console.log(
 		`[pipeline] researcher done in ${elapsed()} — ${brief.length} chars`,
 	);