Back to skill

Security audit

AgentDeals

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a coherent deal-search MCP server, but its hosted service logs user queries and stack details and exposes those logs publicly.

Install only if you are comfortable sending queries and stack details to the hosted AgentDeals service. Avoid entering confidential architecture, business plans, customer names, internal vendor lists, or secrets until the public request-log endpoint is removed or access-controlled and the privacy documentation is corrected.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
src/serve.ts:31589
Finding

Unauthenticated Public Exposure of Persisted MCP Queries and Session Metadata

Content
View full analysis
; user_agent?: string; result_count: number; session_id?: string; client_info?: { name: string; version: string }; } ``` The records are persisted to Upstash Redis when Redis is configured: ```ts export async function logRequest(entry: RequestLogEntry): Promise { const pushed = await redisLpush(REQUEST_LOG_KEY, JSON.stringify(entry)); if (pushed) { // Cap list at REQUEST_LOG_MAX entries await redisLtrim(REQUEST_LOG_KEY, 0, REQUEST_LOG_MAX - 1); } } export async function getRequestLog(limit = 50): Promise { const raw = await redisLrange(REQUEST_LOG_KEY, 0, limit - 1); return raw.map((s) => { try { return JSON.parse(s) as RequestLogEntry; } catch { return null; } }).filter((e): e is RequestLogEntry => e !== null); } ``` Free-form MCP inputs and session identifiers are included in these records. For example ...[truncated 3792 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/check-pricing-changes.js:51
Finding

Server-Side Request Forgery Risk in Pricing Page Monitor

Content
View full analysis
controller.abort(), FETCH_TIMEOUT_MS); try { const res = await fetch(url, { signal: controller.signal, headers: { "User-Agent": "AgentDeals-PricingMonitor/1.0", Accept: "text/html", }, redirect: "follow", }); if (!res.ok) { throw new Error(`HTTP ${res.status}`); } return await res.text(); } finally { clearTimeout(timeout); } } ``` The destination is taken directly from an offer record in `data/index.json`: ```js for (const offer of offers) { if (!offer.url) continue; process.stdout.write(` ${offer.vendor}... `); try { const html = await fetchPage(offer.url); ``` ### Technical Analysis Fetching public vendor pricing pages is necessary for the monitoring feature. However, the code places no security boundary around the destination. It does not validate: - Whether the protocol is `https:`. - Whether the hostname is an approved vendor domain. - Whether DNS resolves to a loopback, private, link-local, multicast, or reserved address. - Whether an IPv6 address targets a local or private network. - Whether redirect destinations remain on an approved public host. - Whether DNS resolution changes between validation and connection. Because `redirect: "follow"` is enabled, even an initially public URL can redirect to an internal service. If an attacker can introduce or modify an offer URL through a malicious contribution, compromised ingestion source, or repository write access, the maintenance pr ...[truncated 1972 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (101)

YARA rule 'agent_skill_credential_exfiltration_webhook': AI agent skill credential harvesting followed by webhook or external exfiltration [agent_skills]

Critical
Category
YARA Match
Confidence
85% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · src/serve.ts (reported line 17)May include surrounding context.

ts
costs.js";
import { recordApiHit, recordSessionConnect, recordSessionDisconnect, recordLandingPageView, getStats, getConnectionStats, loadTelemetry, flushTelemetry, logRequest, getRequestLog, recordPageView, getPageViews } from "./stats.js";
import { openapiSpec } from "./openapi.js";

const __filename = fileURLToPath(import.meta.url);
const __dirname = dirname(__filename);

const PORT = parseInt(process.env.PORT ?? "3000", 10);
const BASE_URL = (process.env.BASE_URL ?? "https://agentdeals.dev").replace(/\/+$/, "");

const INDEXNOW_KEY = process.env.INDEXNOW_KEY ?? "";

const GOOGLE_VERIFICATION_META = process.env.GOOGLE_SITE_VERIFICATION
  ? `<meta name="google-site-verification" content="${process.env.GOOGLE_SITE_VERIFICATION}">\n` : "";

// Patch OpenAPI spec with BASE_URL
openapiSpec.info.contact.url = BASE_URL;
openapiSpec.servers[0].url = BASE_URL;

// Load favicon from logo PNG at startup
const faviconBuffer = readFileSync(join(__dirname, "..", "assets", "logo-400.png"));

// Lo

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · AGENT_README.md (reported line 60)May include surrounding context.

md
Endpoints:
- `POST /mcp` — MCP JSON-RPC requests (requires `Accept: application/json, text/event-stream` header)
- `GET /mcp` — SSE stream for server-initiated notifications
- `DELETE /mcp` — Session termination
- `GET /health` — Health check
- `GET /.well-known/glama.json` — Glama registry ownership verification

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Vendor risk assessment APIs, MCP tool exposure, fuzzy vendor matching, and alternative ranking are broader analytical capabilities than the simple search/compare description suggests. The danger is moderate: they can process user queries in richer ways than disclosed and may expose extra endpoints or inference logic not expected by users.

Content

No source excerpt is available for this finding.

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · assets/og-image.svg (reported line 13)May include surrounding context.

text
</linearGradient>
  </defs>

  <!-- Background -->
  <rect width="1200" height="630" fill="url(#bg)"/>

  <!-- Accent bar at top -->

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · assets/og-image.svg (reported line 45)May include surrounding context.

text
<!-- Tagline -->
  <text x="80" y="370" font-family="system-ui, -apple-system, 'Segoe UI', sans-serif" font-size="30" fill="#94a3b8" font-weight="400">Pricing context for AI agents</text>

  <!-- Stats row -->
  <g transform="translate(80, 440)">
    <rect width="200" height="80" rx="12" fill="white" fill-opacity="0.06"/>
    <text x="100" y="35" font-family="system-ui, sans-serif" font-size="28" fill="#3b82f6" font-weight="700" text-anchor="middle">1,500+</text>