A free Claude Code skill that classifies any cold email reply into 1 of 8 categories and hands you the exact response to send back.
Reply Handling is a Claude Code skill: a markdown playbook that teaches your agent an 8-category taxonomy for cold email replies (meeting interest, answerable question, not right now, already have a solution, wrong person, not interested, out of office, and everything else) plus a ready-made response for each one. Paste in a reply, and it tells you which category it falls into, gives you a human-sounding draft to send back, and tells you whether it needs a human at all or can be handled on autopilot.
The video's claim is that it "classifies every reply into the right category and gives you the exact response, so nothing sits in an inbox waiting on you." The classify-and-draft half of that is fully true and tested. Four invented sample replies (interested, already-have-a-solution objection, not-right-now, wrong-person) were all sorted into the correct category on the first pass, with a usable draft response for each. The "nothing sits in an inbox waiting on you" part needs one honest caveat: this skill does not watch your inbox or send anything by itself. It hands you the category and the draft. Turning that into something that actually fires automatically still means wiring the templates into your sending platform's auto-responder rules, which the skill also gives you step-by-step instructions for, but that's a separate step, not something that happens the moment you install the skill.
One more thing worth knowing: some of the drafts (the "already have a solution" objection, mainly) have a bracket like [specific pain point] that only you can fill in, because only you know your own product. It is not a fully blind copy-paste for every category, most of them are, that one needs ten seconds of thought.
Anyone running cold email who currently lets replies sit for hours because nobody owns "what do I say back." This replaces the mental overhead of deciding, every single time, whether a reply is worth a meeting invite, a polite decline, a referral ask, or nothing at all, and it gives you the actual words to use instead of staring at the inbox.
This copies the whole gtm-skills library onto your machine. You only need one folder out of it.
git clone https://github.com/LeadMagic/gtm-skills.gitThis drops just the reply-handling skill into the folder Claude Code checks for project skills.
mkdir -p .claude/skills
cp -r gtm-skills/skills/outbound/reply-handling .claude/skills/reply-handlingNote: the repo also documents a one-line ./install.sh and a gh skill install command. Both rely on a gh skill extension that was not present on the test machine, and the one-line installer failed outright instead of falling back. The plain copy above is the path that actually worked, and it is also the exact fallback method the repo's own installer uses internally.
Sample input used in testing, an invented cold email reply:
"Hey, actually decent timing, we've been talking about this internally.
Can you send over some times this week to walk me through it?"Abbreviated real output, following the skill's own taxonomy and template:
Category: 1 - Positive, Meeting Interest
Routing: SDR/AE, 15-60 min SLA depending on account tier
Draft reply:
"[First name] - great to hear from you. Here's my calendar, grab a
time that works and I'll confirm: [calendar link]. Looking forward
to it. [Your first name]"Three more invented replies (an "already have a solution" objection, a "not right now" timing reply, and a "wrong person, here's who to talk to" reply) were run the same way. All four landed in the correct category with a usable draft.
./install.sh or gh skill install fails with unknown command "skill" for "gh". Your GitHub CLI does not have the gh skill extension the repo assumes. Skip it and use the plain cp -r copy in Step 2 instead, it lands in the same place.
scripts/cc-gtm.py (the "curated installer") crashes with a SyntaxError about an f-string and a backslash. That script needs Python 3.12 or newer. If you're on an older Python 3, use the cp -r copy method instead.
The draft response still has brackets like [specific pain point] in it. That's expected for objection-style categories. The skill can classify the reply and give you the right structure, it can't know your product's specific pain points for you, fill that one bracket in and send it.
You want it to actually auto-send, not just draft. It won't on its own. Follow the skill's own "Platform Configuration Guide" step to wire the categories and templates into your sending platform's (Smartlead, Instantly, etc.) auto-responder rules.
Built by LeadMagic (https://github.com/LeadMagic/gtm-skills), MIT license, 53 stars at test time. Correction to how the video describes it: it classifies replies and drafts the response correctly, but it does not itself watch an inbox or send anything, the "nothing sits in an inbox waiting on you" part only becomes true once the drafted responses are wired into your sending platform, which is a separate, documented step inside the skill itself.
This is one agent. The 7-Day AI Native Challenge is how you learn to build the next ten yourself, live, with Dany.
Join the Challenge →Too busy to build it yourself? Focentra AI will build, install and train your team on it. Priced on the call.
Talk to Focentra AI