More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    phishing email analyzer

    1

    Triage suspicious emails offline: SPF/DKIM/DMARC, look-alike domains, defanged IOCs, a risk score and a plain-English answer.

    $15

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    phishing email analyzer

    phishing email analyzer

    Example session with this skill installed

    I got an email saying my PayPal account is limited, with an "invoice" attachment and a zip. I saved it as paypal-invoice-phish.eml. Is it real? Don't open anything. Give me a short answer, the evidence, and an IOC list I can send to IT.

    • Read your context and instructions
    • Compiled the phishing email analyzer
    • Generated the data export

    Summary

    • Is this safe? No. HIGH-CONFIDENCE PHISHING, risk 100/100 (29 scored reasons, capped per area).
    • From: "PayPal Service" service@xn--pypal-4ve.test. The domain displays as pаypal.test with a Cyrillic "а", a look-alike of paypal.com (+30).
    • Authentication (top Authentication-Results): SPF softfail, DKIM none,

    DMARC fail with the sender's policy = reject (+25).

    • Reply-To goes to a different domain, paypal-support-desk.test, and the bounce address is mailer-cloud.test.
    • Route: 3 hops, origin IP 203.0.113.45. Hop 2 is stamped 6 minutes before hop 1, a sign of clock skew or a forged Received line.
    • Links (4, none opened): link text says www.paypal.com but goes to paypal.com.account-verify.test. A Microsoft Safe Links wrapper was unwrapped to hxxp://203[.]0[.]113[.]80:8080/invoice/view.php.
    • Attachments (2, hashed in memory): Invoice_88231.pdf.html is a 281-byte HTML login page with a password field that posts to account-verify.test (sha256 df70ae9e…a1589b). Remittance_Advice.zip hides Payment_Details.pdf.js, a file type that Outlook and Gmail both block (+35).

    Deliverable

    • answer.txt: a plain-English verdict, 4 reasons and what to do
    • triage.md: full report with every reason and its points
    • iocs.csv: 24 defanged indicators (3 email, 5 domain, 3 IP, 4 URL, 3 filenames, plus SHA-256/SHA-1/MD5 for both attachments)

    Next steps

    1. Don't click, open, reply or call any number in it. Use PayPal's real site or app if you're worried.
    2. Report it with your mail app's "Report phishing" button, forward it to reportphishing@apwg.org and report it at ReportFraud.ftc.gov. Then delete it.
    3. Send iocs.csv to IT so they can block the domains, IPs and hashes.
    4. If you already typed a password, change it everywhere you reused it and turn on MFA.

    (Numbers come from a real run on the bundled sample. Its domains are fictional .test names.)

    Invoice_88231.pdf

    CSV · data export

    Generated

    Example file from a real run - the skill writes it into your workspace.

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

    About this skill

    Know if an email is phishing without clicking anything.

    phishing email analyzer gives your AI agent a SOC-style triage script for suspicious email. Hand it a saved .eml file or the raw headers you copied out of Gmail or Outlook. It reads the evidence the receiving mail server recorded, checks every link, sender address and attachment against tested rules, and returns a 0-100 risk score that lists each reason with its points. You also get a plain-English "Is this safe?" answer for the person who received the email, plus a defanged IOC file for whoever blocks it. Everything runs locally. Links are never opened and attachments are never saved or run.

    What it does

    • Authentication: reads SPF, DKIM and DMARC results from Authentication-Results (RFC 8601), Received-SPF and Microsoft's compauth, checks each one's alignment with the From domain, and reports the sender's DMARC policy. It uses only the top (your provider's) Authentication-Results header, because lower ones can be forged.
    • Route forensics: reads the Received chain oldest-first and reports each hop's IP class (private, CGNAT, documentation, public), the origin IP, hop delays, and negative time jumps that point to a forged or skewed line.
    • Sender checks: compares From, Reply-To and Return-Path, catches display-name spoofing ("PayPal Service" sending from some other domain), freemail senders using brand lures, and Microsoft's own phishing/spoof verdicts (X-Forefront-Antispam-Report CAT codes).
    • Look-alike domains: punycode (xn--), mixed Cyrillic/Greek/Latin letters, homoglyphs based on Unicode TR39 confusables (1,586 mappings bundled), 1-2 letter typos, brand-secure.com and paypal.com.evil.net patterns. Covers 30 common brands, and you can add your own brands and your company's domain to catch CEO and vendor fraud (BEC).
    • Links: pulls URLs from text, HTML anchors, forms, iframes and meta refresh. Flags link text that shows one domain but points to another, bare IPs, odd ports, URL shorteners and http logins. Unwraps Microsoft Safe Links so you see the real target.
    • Attachments: hashes each file in memory (MD5, SHA-1, SHA-256), detects the real file type from its bytes, flags double extensions, HTML/SVG login pages, password fields, forms that post elsewhere, macro-enabled Office files, encrypted archives, and file types on Outlook's and Gmail's block lists. Zips are listed but never extracted.
    • Outputs: a terminal report, triage.md, iocs.csv (all defanged and safe from CSV formula injection), answer.txt in plain language, and triage.json.

    Problems it solves

    • "Is this email real?" questions from family, staff or clients, answered with evidence instead of a gut feeling
    • CEO gift-card and fake-vendor emails that pass SPF, DKIM and DMARC because the attacker owns a look-alike domain
    • Raw headers too long and confusing to read by hand
    • Pasting a suspicious link into a browser or scanner "just to check"
    • Rebuilding IOC lists by hand for a blocklist, ticket or abuse report

    Who it's for

    Small-business owners and office managers, MSPs and IT generalists, help-desk and SOC analysts who want a fast first pass, and anyone who gets asked "is this a scam?". The plain-English answer works for non-technical people, and the scored report and IOCs work for analysts.

    What's included

    • SKILL.md: safety rules for the agent, how to capture headers in Gmail and Outlook, run commands, how to read each signal, and the order to deliver results
    • scripts/phish_triage.py (main CLI), eml_parse.py, indicators.py, make_samples.py (4 harmless test emails using reserved .test domains and RFC 5737 IPs)
    • scripts/data/: Unicode confusables subset, brand domain list, and blocked/risky file-type lists with their sources
    • references/: every standard and provider rule with its official source (RFCs, Microsoft, Google, FTC, CISA, Unicode), the full scoring table, and a header-forensics guide
    • examples/expected-results.md

    Standard-library Python only. No API keys, no network calls, no installs.

    • Offline only: no DNS, WHOIS or domain-age lookups, reputation feeds or sandboxing. DKIM signatures aren't re-verified cryptographically. The skill trusts the result your mail provider recorded.
    • Organizational-domain matching is an offline approximation of the DNS tree walk in RFC 9989 (DMARCbis).
    • The brand list is a starter set of 30 brands. Add others with --brand, and pass your company domain with --org-domain.
    • Wording checks (urgency, gift cards, "verify your account") are English-only and carry few points. Non-English scams lean on the technical signals.
    • QR codes inside images and links inside PDF or Office attachments aren't read.
    • Outlook .msg files aren't supported, and some Outlook "Save as EML" exports drop the original headers. The skill tells you to use "View message details" in that case.
    • A low score isn't proof of safety. A well-written email from a hacked real account can look clean. Always verify money or password requests another way.

    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 any payment.

    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

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    Requires

    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    File Scopes

    **/*.eml
    **/*.txt
    ./*.md
    ./*.txt
    ./*.csv
    ./*.json
    ./samples/*

    All analysis is local and offline. Links are only parsed as text and are never fetched. Attachments are decoded in memory for hashing and type detection, and are never written to disk, extracted or run. No credentials, no mailbox access, no package installs, no system changes. The scripts don't import networking or subprocess modules, and they're plain, commented Python with no obfuscation.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    The analyzer works by inspecting the raw code of an email (.eml file or copied headers). It looks for authentication failures (SPF/DKIM/DMARC), traces the sender's path, checks for look-alike domains using Unicode homoglyph detection, and flags dangerous attachment types without ever opening them.

    This skill is a standalone Python-based toolkit designed to be used by any AI agent with file-system or shell access. It uses only the Python standard library, requiring no external dependencies, API keys, or internet connectivity to function.

    No, the skill is designed for maximum safety. It performs all analysis offline, zips are never extracted, and all extracted links/IOCs are 'defanged' (e.g., changing http to hxxp) so they cannot be accidentally clicked or executed during the triage process.

    You receive the core triage and parsing scripts, a comprehensive data library of brand domains and Unicode confusables, and a detailed safety guide. It also includes sample 'harmless' phishing emails so you can test your agent's reporting capabilities safely.

    Yes, the included data files for brands and blocked file types are user-editable. You can easily add your own company domains, specific executive names to watch for, or unique vendor domains to the 'brands' list to better detect Business Email Compromise (BEC) attacks.

    The analyzer provides a multi-layered output: a technical JSON/Markdown report for IT professionals, a CSV file of indicators for blocking, and a specific 'plain-English' summary designed to be read directly to the end-user who received the suspicious email.

    New in Security