fix: Discord attachment download support

[changhansung]

Summary

Fixes #2 — Discord image attachments were received as blank messages.

Changes (plugins/agend-plugin-discord/src/discord-adapter.ts)

• Store attachment URLs: Added attachmentUrls Map to cache Discord CDN URLs during message handling
• Detect image attachments: Set kind: 'photo' for image/* content types so processAttachments() auto-downloads them (previously hardcoded to 'document')
• Implement downloadAttachment: Fetch from stored CDN URL → write to inboxDir → return local path (matching Telegram adapter pattern)
• Prevent memory leak: Clean up URL from map after successful download

Before

• Discord images arrived as blank messages — downloadAttachment() threw not yet implemented
• All attachments were tagged as document, so image auto-download never triggered

After

• Images correctly identified and auto-downloaded via processAttachments()
• Agents receive image_path metadata and can read the image directly
• Behavior matches Telegram adapter

[amazon-q-developer]

Summary

This PR successfully implements Discord attachment download support, enabling image attachments to be processed correctly. However, there are 3 critical issues that must be addressed before merge:

Critical Issues Found

1. Memory Leak Risk (Line 58): Attachment URLs accumulate indefinitely when downloads don't occur or fail
2. File Collision Risk (Lines 379-380): Unsanitized filenames can cause overwrites and data loss
3. Resource Leak (Lines 382-388): Failed downloads leave URLs in memory due to missing cleanup guarantees

Required Actions

All three issues have blocking severity and require fixes. Code suggestions have been provided for issues #2 and suzuke#3. Issue #1 requires implementing a cleanup strategy (TTL-based expiration or size-based eviction).

Once these are resolved, the implementation will correctly match the Telegram adapter pattern as intended.

---

You can now have the agent implement changes and create commits directly on your pull request's source branch. Simply comment with /q followed by your request in natural language to ask the agent to make changes.

[amazon-q-developer]

🛑 Memory Leak Risk: Attachment URLs are stored indefinitely in the Map but only deleted after successful download. If downloadAttachment is never called (e.g., for non-photo attachments) or download fails, URLs accumulate in memory over the bot's lifetime. Consider implementing a cleanup mechanism such as TTL-based expiration or size-based eviction to prevent unbounded memory growth.

[amazon-q-developer]

🛑 File Collision Risk: Using unsanitized filenames from Discord CDN URLs can cause overwrites when multiple attachments share the same name. Add unique identifier (timestamp or fileId) to prevent data loss from concurrent or sequential attachments with identical names.

Suggested change

	const filename = new URL(url).pathname.split("/").pop() ?? fileId;
	const localPath = join(this.inboxDir, filename);
	const filename = new URL(url).pathname.split("/").pop() ?? fileId;
	const localPath = join(this.inboxDir, `${fileId}-${filename}`);

[amazon-q-developer]

🛑 Resource Leak: If pipeline fails after fetch succeeds, the URL remains in the Map permanently because cleanup only executes after successful pipeline completion. Wrap cleanup in try-finally to ensure deletion regardless of download outcome.

Suggested change

	const response = await fetch(url);
	if (!response.ok) throw new Error(`Failed to download attachment: ${response.status}`);
	if (!response.body) throw new Error("No response body");
	await pipeline(Readable.fromWeb(response.body as import("stream/web").ReadableStream), createWriteStream(localPath));
	this.attachmentUrls.delete(fileId);
	return localPath;
	const response = await fetch(url);
	if (!response.ok) throw new Error(`Failed to download attachment: ${response.status}`);
	if (!response.body) throw new Error("No response body");
	try {
	await pipeline(Readable.fromWeb(response.body as import("stream/web").ReadableStream), createWriteStream(localPath));
	return localPath;
	} finally {
	this.attachmentUrls.delete(fileId);
	}