More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    FlowSentinel Network Investigator

    1

    Let your AI agent investigate suspicious network flows: map traffic to local processes, read explainable scores and MITRE ATT&CK techniques.

    Free

    Security scanned
    FlowSentinel Network Investigator

    FlowSentinel Network Investigator

    Example session with this skill installed

    Analyze sample.pcap and show me the most suspicious flows. Explain why they are flagged.

    • Read your context and instructions
    • Compiled the flowsentinel network investigator

    Real output of analyze_pcap on a generated test capture (synthetic beaconing scenario, not a real incident).

    Summary

    46 flows in 138 packets. Risk summary: 0 critical, 20 high, 0 medium, 26 low.

    Most suspicious flow

    FieldValue
    Flow192.168.1.50:50019 -> 198.51.100.23:4444 (TCP)
    Score6.8 / 10 (HIGH)
    MITRE ATT&CKT1583 Acquire Infrastructure

    Why it is flagged

    • known high-risk port 4444 (Metasploit default listener)
    • no reverse DNS: direct IP connection
    • destination IP on a threat-intel blocklist

    The other 19 high-risk flows have the same destination, the same port and a fixed 30-second spacing, which is the classic shape of

    beaconing.

    Next steps

    1. Live case: run live_watch on the owning process, then analyze_process for its parent chain.
    2. Run scan_process to check the binary hash and its location.
    3. Use get_flow_history for 198.51.100.23 to see how long this has been going on.

    Scores are heuristics, not verdicts: verify before taking action.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Which process is sending data to that unknown IP? This skill teaches your AI agent to answer that question with the free, open-source MCP-FlowSentinel server: it captures traffic locally, attributes every flow to the process that owns the socket, and scores flows with explainable behavioral signals mapped to MITRE ATT&CK.

    What it does

    • Picks the right tool for you: lists interfaces, then runs a short capture (analyze_network), a targeted watch on one process or IP (live_watch), a replay of a saved .pcap/.pcapng (analyze_pcap) or a look back at history (get_flow_history).
    • Drills into suspects: parent chain, open connections, GeoIP and ASN (analyze_process), binary SHA-256, location check and optional VirusTotal hash lookup (scan_process).
    • Explains instead of accusing: every flagged flow comes with its score (0-10), the human-readable reasons, the owning process and destination, and the MITRE ATT&CK techniques.

    Built-in guardrails

    • Starts with short captures and filters with min_score / top_n so you are not buried in raw flows.
    • Never claims a host is compromised from a single score; recommends corroborating checks.
    • Encrypted payloads are never decrypted; conclusions rely on metadata, JA3/JA3S/HASSH fingerprints, timing and process context, and the skill says so.
    • Never asks you to paste API keys into the chat.

    Requirements

    The mcp-flowsentinel server (Go, MIT licensed, https://github.com/ClementG91/MCP-FlowSentinel) registered in your agent. Live capture needs libpcap (Linux/macOS) or Npcap (Windows) and capture privileges; offline pcap analysis needs neither.

    Who it is for

    Security-minded developers, SOC analysts, incident responders and homelab owners who want an AI assistant that works from real local evidence. Scores are heuristics, not verdicts: always verify before acting.

    Built by a developer behind several open-source security tools (FlowSentinel, WinSight).

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.

    Requires libpcap/Npcap and capture privileges. No payload decryption. Unattributed flows occur with short-lived sockets or containers. Binary scans require user-configured VirusTotal API key.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 23 days ago

    • Free with an Agensi account

    Software
    FlowSentinel MCP
    libpcap
    orNpcap

    Get the server from github.com/ClementG91/MCP-FlowSentinel. Use libpcap on Linux or macOS, Npcap on Windows.

    Permissions

    Network Access
    Listed23 days ago

    What's inside

    Frequently Asked Questions

    FlowSentinel bridges the gap between raw network traffic and your operating system's process list, allowing you to instantly identify exactly which binary or PID is responsible for a specific network connection—a task that typically requires manual, time-consuming forensic correlation.

    This skill is designed for any AI agent or platform that supports the Model Context Protocol (MCP) and requires the FlowSentinel server (stdio) to be running on the host machine to execute network calls.

    To perform live captures and link sockets to local processes, the underlying FlowSentinel server requires elevated (Administrator/Root) privileges and the installation of libpcap (Linux/macOS) or Npcap (Windows).

    While FlowSentinel cannot see inside encrypted payloads (like TLS), it uses JA3/HASSH fingerprinting and behavioral heuristics to identify suspicious patterns, C2 communication, and beaconing based on connection metadata.

    The skill itself is free. It tells your agent how to use the free, open-source mcp-flowsentinel MCP server (MIT licensed, https://github.com/ClementG91/MCP-FlowSentinel). Install and register that server in your agent, plus libpcap (Linux/macOS) or Npcap (Windows) if you want live capture; analyzing saved pcap files needs neither.

    Yes, the skill includes an 'analyze_pcap' function that allows you to upload existing capture files for the AI to parse, explain, and score based on its integrated forensic heuristics.

    New in LLM Ops

    Free