More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Support Reply Library: Saved Replies from Your Real Tickets

    1

    The agent drafts a support library for your team to review and paste into a helpdesk: saved replies, help articles, a tone guide, an escalation map and a coverage report.

    $4

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Support Reply Library: Saved Replies from Your Real Tickets

    Support Reply Library: Saved Replies from Your Real Tickets

    Example session with this skill installed

    Build a support reply library from facts.md and tickets.csv. Email is the main channel, with a short version of each reply for chat. good-replies.md has two replies we like.

    • Read your context and instructions
    • Compiled the support reply library

    R01. Will the iPhone 15 case fit an iPhone 15 Pro?

    • Use when: someone asks before buying whether an iPhone 15 case fits the 15 Pro (or the other way round).
    • Check first: which model they have: 15 or 15 Pro.
    • Facts used: F26

    Reply

    Hi {first_name},

    The iPhone 15 case won't fit the 15 Pro: the camera cutout is different. You need the case listed for the
    iPhone 15 Pro.

    Each product page lists the exact models that case fits, so it's worth a quick look before you order.

    {agent_name}

    Short

    It won't fit, the camera cutout on the 15 Pro is different. Go for the case listed for the 15 Pro; each product page shows the exact models it fits.

    R02. Which model does this fit? (answer is on the product page)

    • Use when: a fit question about a case or screen protector for a named phone (for example "which screen
      protector for a Pixel 8a"). For iPhone 15 against 15 Pro use R01.
    • Check first: the product page for {item_name}: is the customer's model in its list? If you cannot confirm it
      from the page, do not guess; ask Sam.
    • Facts used: F26, F29

    Reply

    Hi {first_name},

    For the {phone_model} you need {item_name}: it's listed for that model on its product page.

    If you're ever unsure, the list of models on each product page is the one to trust.

    {agent_name}

    Short

    For the {phone_model} you want {item_name}, it's listed for that model on the product page.

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

    About this skill

    English

    For shop owners and small support teams who answer the same questions every day.

    Give your agent what your business already knows (policies, shipping, pricing, known issues) and a pile of past customer messages. The agent drafts a support library for your team to review and paste into a helpdesk: saved replies, help articles, a tone guide, an escalation map and a coverage report.

    What it writes each run

    • saved-replies.md: 20 to 40 replies grouped by topic. Each has when to use it, the one thing to check before sending, the fact numbers it rests on, the full reply with {placeholders}, and a short variant for chat or WhatsApp.
    • help-articles/: 5 to 10 task-shaped articles for questions customers can solve alone.
    • tone-guide.md: voice rules, phrases to use and avoid, how to decline, how to apologise once and move to the fix.
    • escalation-map.md: which cases go to whom, what to pass along, and which reply to send meanwhile.
    • coverage.md: ticket themes with counts, the replies and articles that cover each, and the gaps found, each with the question to ask the business.

    Only what the business has stated

    Each reply cites numbered facts from your material. The skill instructs the agent to add no policy, delivery time, compensation or cause, and to give a theme with no rule a holding reply that promises only who follows up, marked [NEEDS: ...].

    Themes counted from your tickets

    A bundled Python script (standard library, no network calls) counts tickets per theme from a keyword map the agent writes after reading the tickets. It reads CSV, JSON Lines or plain text and normalises Arabic spelling variants. The counts are approximate, and the coverage file notes where reading a ticket puts it elsewhere.

    Difficult cases included

    Angry customer, late delivery, damaged item, billing dispute, account access, a request outside policy, and a holding reply for legal threats that escalates. Declines say what is possible first.

    Honest about its limits

    Thin facts give many holding replies. The files are Markdown; nothing connects to your helpdesk or reaches customers. The quality check is the agent's own, so have a second person read the replies before use.

    Arabic

    On request the library is written in Arabic, Modern Standard by default or a dialect that follows your sample replies; placeholders stay in Latin letters. Arabic output was tried on one test case and was not reviewed by a second native reader; check it before sharing.

    In the download: SKILL.md, the script, three reference files (method, difficult cases, Arabic notes), templates, and one complete example: a fictional phone accessories shop with 80 tickets in, and 37 replies, seven help articles, a tone guide, an escalation map and a coverage report out.

    How it was tried: Tried on two cases (an English phone accessories shop, an Arabic perfume shop); the download includes one worked example in English. The bundled script has 15 unit tests.

    بالعربي

    لأصحاب المتاجر وفرق الدعم الصغيرة التي تجيب عن الأسئلة نفسها كل يوم.

    أعطِ وكيلك ما يعرفه مشروعك (السياسات، والشحن، والأسعار، والمشكلات المعروفة) ومجموعة من رسائل الزبائن السابقة. تستلم مكتبة دعم جاهزة للّصق في نظام الدعم.

    ماذا تستلم

    • ملف saved-replies.md: من 20 إلى 40 ردًّا مرتبة بحسب الموضوع، ولكل رد: متى يُستخدم، وما الذي تتحقق منه قبل الإرسال، والنص الكامل، ونسخة مختصرة للمحادثة أو واتساب.
    • مجلد help-articles/: من 5 إلى 10 مقالات مساعدة.
    • ملف tone-guide.md لأسلوب الكتابة، وملف escalation-map.md لتحويل الحالات.
    • ملف coverage.md: مواضيع الرسائل بأعدادها، وما يغطي كل موضوع، وكل نقص مع السؤال الذي يُطرح على صاحب المشروع.

    ما قاله المشروع فقط

    كل رد يستند إلى حقائق مرقّمة من موادك. تنص التعليمات على ألا تُضاف سياسة ولا موعد ولا تعويض. والموضوع الذي لا قاعدة له يأخذ ردًّا مؤقتًا بلا وعود ويُعلَّم بصيغة [NEEDS: ...].

    حدود واضحة

    أعداد المواضيع يحسبها سكربت بايثون محلي بالكلمات المفتاحية، فهي تقريبية. والملفات بصيغة Markdown ولا يُرسل شيء إلى الزبائن. ويُستحسن أن يقرأ الردود شخص ثانٍ قبل استعمالها.

    العربية

    عند الطلب تُكتب المكتبة بالعربية الفصحى، أو بلهجة تتبع نماذج ردودك، وتبقى الرموز بين القوسين بحروف لاتينية.

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

    The library is only as good as the facts you give: thin facts produce many holding replies and a long [NEEDS: ...] list, and it does not write or change your policies. Theme counts are keyword-based and approximate; a ticket can be filed under a neighbouring theme, and long multi-topic messages misfile more often. The coverage file notes the cases found by reading. The script does not detect language or split multi-question tickets, and plain-text input needs one ticket per line or blank-line blocks. Output is Markdown only: no helpdesk import was tried and nothing is sent to customers. The quality check is the agent's own self-check, so a second reader is advisable before the replies go live. Nobody outside the build has reviewed the example replies; the Arabic dialect variants follow your sample replies and were not read by a second native reader, and only Modern Standard and Iraqi were exercised in a test case. الردود مبنية على معلوماتك فقط، وتحتاج قراءة ثانية قبل استعمالها.

    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

    Money-back guarantee

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

    Trust & safety

    Security scan

    Security scan passed 3 days ago

    Software
    python3 (3.9 or newer) for the counting script

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Reads the facts, tickets and sample replies you point it at and does not change them. Writes saved-replies.md, a help-articles folder, tone-guide.md, escalation-map.md, coverage.md and a work folder with the theme map and count tables into a folder in your project (default support-library). Runs its bundled Python script locally (ticket_themes.py); it uses the standard library only, makes no network calls and writes only into the folder you name. The skill instructs the agent not to copy customer names, order numbers or addresses from tickets into the library; check before sharing. يقرأ ملفاتك ويكتب ملفات المكتبة فقط، ولا يتصل بنظام الدعم.

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    From your own facts. Each reply lists the fact numbers it rests on, and the skill instructs the agent to add no policy, time, compensation or cause. Read the replies against your policies before use. من معلوماتك أنت، وكل رد يذكر أرقام الحقائق التي يستند إليها.

    The agent asks you about it. If no answer comes, the theme gets a holding reply that promises only who follows up, and a [NEEDS: ...] line with the question to ask the business. يسألك عنه، وإلا يكتب ردًّا مؤقتًا بلا وعود ويعلّم النقص.

    They are keyword counts from the script, so approximate. The agent also reads the tickets and notes where its reading differs, without changing the script's numbers. تقريبية لأنها بالكلمات المفتاحية، مع ملاحظات من قراءة الوكيل.

    It says what is possible first: the request is described as outside the stated condition, such as the return window, followed by what can be offered instead or who will review it. يذكر الشرط الذي خرج عنه الطلب ثم ما يمكن تقديمه بدلًا منه.

    No. It writes Markdown files for you to paste in. Nothing is sent to customers. لا، تستلم ملفات Markdown تلصقها بنفسك.

    The skill instructs the agent not to: everything specific to a case is to be a placeholder such as {first_name} or {order_number}, listed at the end of the file. Check the files for customer details before use. التعليمات تمنع ذلك، وكل ما يخص الحالة رمز بين قوسين.