Claude 24f816b6a3
Consolidate 22 sibling repos into layered organism structure
Place useful parts of the surrounding repos into sica-fondt by layer, per the
body model (Ada = membrane; brain/endocrine/capabilities/knowledge non-Ada):

- brain/        LLM reasoning + providers (dapr, hermes, MoMoA)
- capabilities/ REPRAG sidecars: hermes tools/skills, dapr tools, parallel
                dispatch, A51 channels, and the OSINT cluster
- knowledge/    LORAG corpus: 754 cyber-skills, agency personas, secure-coding,
                MITRE ATT&CK data
- reference/    defensive threat-reference (C3, shhbruh doc) + AdaYaml parser

License handling: AGPL sources (worldosint, advanced_evolution, mercury,
Reticulum) and GPL DeTTECT are SPEC-only clean-room/port descriptions — no
copyleft code copied. MIT/Apache/data parts copied as working trees.

Safety: shhbruh escape/persistence material and C3 covert-C2 kept as reference
only, not wired into the running organism. See CONSOLIDATION.md.

https://claude.ai/code/session_01UehUqEXXJJCsHoA4voCU5c
2026-06-10 06:53:01 +00:00

291 lines
10 KiB
TypeScript

/**
* Copyright 2026 Reto Meier
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
import { MultiAgentTool } from '../multiAgentTool.js';
import { MultiAgentToolContext, MultiAgentToolResult, ToolParsingResult } from '../../momoa_core/types.js';
import { getAssetString } from '../../services/promptManager.js';
import * as path from 'node:path';
import { Buffer } from 'node:buffer';
import { addDynamicallyRelevantFile, updateFileEntry } from '../../utils/fileAnalysis.js';
const LARGE_FILE_LIMIT_KB = 100;
const cache = new Map<string, MultiAgentToolResult>();
/**
* Implements the URL Fetch Tool, providing functionality to fetch content
* from a given URL using the 'fetch' API.
* * This tool now automatically saves downloaded files to the project context.
* * Large text files are saved as "binary" to hide them from the diff, but a snippet is returned.
*/
export const urlFetchTool: MultiAgentTool = {
displayName: "URL Fetcher",
name: 'URL/FETCH{',
endToken: '}',
/**
* Executes the URL fetch tool.
* @param params The parameters for the tool's execution, expecting a 'url' property.
* @param context The ToolContext object containing necessary runtime information.
* @returns A promise that resolves to the URL's content or an error message.
*/
async execute(params: Record<string, string>, context: MultiAgentToolContext): Promise<MultiAgentToolResult> {
const providedUrl = params.url;
if (!providedUrl) {
return {
result: `Error: 'url' parameter is missing for ${this.displayName} tool.`
};
}
const url = providedUrl.trim();
if (!url.startsWith('http://') && !url.startsWith('https://')) {
return {
result: `Error: Invalid URL. Must start with 'http://' or 'https://'.`
};
}
// Check cache (Only for things previously determined to be pure text content/webpages)
if (cache.has(url)) {
context.sendMessage({
type: 'PROGRESS_UPDATE',
message: `Using cached content from \`${url}\``,
}
);
return cache.get(url)!;
}
// Cache Miss: Proceed with fetch
context.sendMessage({
type: 'PROGRESS_UPDATE',
message: `Fetching content from \`${url}\``,
}
);
try {
const response = await fetch(url);
// This block handles HTTP errors (e.g., 404, 503) - DO NOT CACHE
if (!response.ok) {
return {
result: `Error: Failed to fetch URL '${url}'. Server responded with HTTP Status: ${response.status} (${response.statusText})`
};
}
// --- FILE DOWNLOAD & CONTENT ANALYSIS ---
// 1. Determine Filename
let filename = '';
const contentDisposition = response.headers.get('content-disposition');
if (contentDisposition) {
const match = contentDisposition.match(/filename="?([^"]+)"?/);
if (match) filename = match[1];
}
if (!filename) {
try {
const urlObj = new URL(url);
filename = path.basename(urlObj.pathname);
} catch (e) {
// Ignore URL parsing errors
}
}
// 2. Fetch Content & Analyze Type
const buffer = Buffer.from(await response.arrayBuffer());
const contentType = (response.headers.get('content-type') || '').toLowerCase();
// Heuristic: Is this text? (Includes JSON, XML, JS, TS, HTML, plain text)
const isText = contentType.includes('text/') ||
contentType.includes('json') ||
contentType.includes('xml') ||
contentType.includes('javascript') ||
contentType.includes('typescript') ||
contentType.includes('application/x-sh');
// Heuristic: Is this a file download or a webpage?
const isHtml = contentType.includes('text/html');
// const hasFileExtension = filename.includes('.') && !filename.match(/\.(html|htm|php|asp|jsp)$/i);
const isLarge = buffer.length > LARGE_FILE_LIMIT_KB * 1024;
// We save as a file if:
// A. It has a file extension AND is not a generic webpage
// B. OR it is explicitly not text (Binary)
// C. OR it is text but "Large" (automatically treated as a file asset)
const shouldSaveAsFile = (filename && !isHtml) || !isText || isLarge;
if (shouldSaveAsFile && filename) {
// --- SAVE TO PROJECT CONTEXT ---
let resultMessage = '';
const analysisDescription = `[Downloaded Asset] Content fetched from ${url}`;
// Case A: Small Text File -> Save to fileMap (Visible in Diff)
if (isText && !isLarge) {
const content = buffer.toString('utf-8');
// Ensure no conflict in binary map
if (context.binaryFileMap.has(filename)) context.binaryFileMap.delete(filename);
context.fileMap.set(filename, content);
context.editedFilesSet.add(filename);
addDynamicallyRelevantFile(filename);
await updateFileEntry(filename, context.fileMap, undefined, {
filename,
description: analysisDescription,
relatedFiles: ''
});
context.sendMessage({
type: 'PROGRESS_UPDATE',
message: `Successfully downloaded \`${filename}\` (${buffer.length} bytes).`,
}
);
return {
result: `Successfully downloaded '${filename}' (${buffer.length} bytes).\nSaved to project files (Visible in Diff).\n\n--- Content ---\n${content}`,
transcriptReplacementID: url,
transcriptReplacementString: `--- Downloaded file '${filename}' ---`
};
}
// Case B: Large Text OR Binary -> Save to binaryFileMap (Hidden from Diff)
// We use binaryFileMap for large text files to prevent them from flooding the LLM context via the Diff.
const base64Content = buffer.toString('base64');
// Ensure no conflict in text map
if (context.fileMap.has(filename)) context.fileMap.delete(filename);
context.binaryFileMap.set(filename, base64Content);
context.editedFilesSet.add(filename);
addDynamicallyRelevantFile(filename);
await updateFileEntry(filename, context.fileMap, undefined, {
filename,
description: isText ? `${analysisDescription} (Large Text)` : `${analysisDescription} (Binary)`,
relatedFiles: ''
});
if (isText) {
const snippet = buffer.subarray(0, LARGE_FILE_LIMIT_KB*1000).toString('utf-8');
resultMessage = `Successfully downloaded '${filename}' (${buffer.length} bytes).\n` +
`Saved to project files as a large asset (Hidden from Diff to save context).\n\n` +
`--- First ${LARGE_FILE_LIMIT_KB}KB Snippet of '${filename}' ---\n${snippet}\n\n... (remaining content saved to file) ...`;
} else {
resultMessage = `Successfully downloaded '${filename}' (${buffer.length} bytes).\n` +
`Saved to project files (Binary).\n[Binary content hidden]`;
}
context.sendMessage({
type: 'PROGRESS_UPDATE',
message: `Successfully downloaded '${filename}' (${buffer.length} bytes).`,
}
);
return {
result: resultMessage,
transcriptReplacementID: url,
transcriptReplacementString: `--- Downloaded large/binary file '${filename}' ---`
};
}
// --- FALLBACK: GENERIC WEBPAGE CONTENT ---
// (This path is taken for small HTML pages or content without filenames, usually just for reading)
const prefix = await getAssetString('url-content-prefix');
const suffix = await getAssetString('url-content-suffix');
const replacementString = await getAssetString('url-content-removed');
const content = buffer.toString('utf-8');
context.sendMessage({
type: 'PROGRESS_UPDATE',
message: `\`\`\`\`\n${content}\n\`\`\`\``,
}
);
const toolResult: MultiAgentToolResult = {
result: `${prefix}\n${content}\n${suffix}`,
transcriptReplacementID: url,
transcriptReplacementString: `${prefix}\n${replacementString}\n${suffix}`
};
cache.set(url, toolResult);
return toolResult;
} catch (error) {
// This block handles network errors (e.g., DNS failure, connection refused) - DO NOT CACHE
const errorMessage = error instanceof Error ? error.message : String(error);
context.sendMessage({
type: 'PROGRESS_UPDATE',
message: errorMessage,
});
return {
result: `Error: Network failure while fetching URL '${url}'. Details: ${errorMessage}`
};
}
},
/**
* Extract parameters from the tool invocation string.
* @param invocation The string used to invoke the tool.
* @returns The parameter names and corresponding values.
*/
async extractParameters(invocation: string, _context: MultiAgentToolContext): Promise<ToolParsingResult> {
if (invocation.trim().endsWith("}")) {
const url = invocation.trim().slice(0, -1).trim();
return {
success: true,
params: {
url
}
};
} else {
return {
success: false,
error: `Invalid syntax for the ${this.displayName} Tool. Make sure you include the curly brackets.`
}
}
}
};
/**
* Generates a formatted string of the current cache contents.
* @returns A string listing all cached URLs and their content.
*/
export function getFormattedCacheContents(): string {
const entries: string[] = [];
for (const [url, toolResult] of cache.entries()) {
// The 'result' property holds the full content, including prefixes/suffixes,
// which is what was cached.
const content = toolResult.result;
const entryString = [
`URL: ${url}`,
`Content:`,
'```',
content,
'```'
].join('\n');
entries.push(entryString);
}
// Join all entries with the "----" separator
return entries.join('\n----\n');
}