Cold Outreach Sequence
Builds a 5-touch LinkedIn + email cold outreach sequence with personalization tiers
Testad · Fungerar
Vad den gör
Generates a full cold outreach sequence for a named prospect: LinkedIn connection request (300-char cap enforced), first message, two follow-ups, and a break-up message on a Day 0/7/14/21 cadence, plus a personalization tier assignment, a pipeline tracking table, and a self-critique pass. Triggers when you ask for cold outreach, prospect messaging, a LinkedIn connection request, a cold email cadence, or a follow-up sequence for leads you have not spoken to yet. Has quick, standard, and deep modes for one message, one full sequence, or a multi-prospect campaign system.
Testrapport
Cloned the repo (GitHub API was rate-limited), found SKILL.md at cold-outreach-sequence/, installed it into a throwaway HOME with cp -R, and confirmed valid name+description frontmatter; raw-fetched three paths (SKILL.md, SKILL-OC.md, positioning-basics/SKILL.md — all HTTP 200), and grep found no curl|sh, base64, credential reads, or injection text. Trigger phrasings tested — SHOULD fire: "Write me a LinkedIn connection request and follow-up sequence for a VP of Eng at a Series A fintech", "I've got 40 cold leads from a conference, help me build an outreach cadence that doesn't feel spammy", "Draft a 4-touch cold email sequence to book meetings with SaaS founders"; SHOULD NOT: "Write a nurture email sequence for people who already downloaded our whitepaper" (warm inbound, not cold prospecting), "Optimize my LinkedIn profile headline and about section" (profile copy, not outreach) — 5/5 correct. For output I wrote two artifacts for the same prospect brief (/tmp/coseq-eval/baseline.md and /tmp/coseq-eval/skill.md): my baseline connection request measured 454 characters, which LinkedIn's 300-char connect note would reject outright, and it pitched positioning plus the "20-30% lift" claim inside the connect note; the skill-followed one came in at 168 chars with no pitch. The baseline follow-up was literally "following up on my note below in case it got buried" with zero new information — the exact thing the skill body forbids — while the skill version added a new angle, and the skill run also produced a Day 21 break-up message, a Day 201 resurface date, a Tier 1 assignment, a pipeline table, and a self-critique that caught my first message ending in a meeting ask and rewrote it as a soft ownership question. Docs lose a point for README hype ("Your agent applies it → Every time. Perfectly.") and an install snippet that dumps README/CHANGELOG/scripts into ~/.claude/skills.
Testad: 2026-07-21 · Claude Code 2.x (agent harness)
Installation
git clone --depth 1 https://github.com/BrianRWagner/ai-marketing-claude-code-skills.git /tmp/cold-outreach-sequence-src mkdir -p ~/.claude/skills cp -R /tmp/cold-outreach-sequence-src/cold-outreach-sequence ~/.claude/skills/cold-outreach-sequence # No dependencies, no API keys, no scripts — SKILL.md is pure prose instructions. # The body asks for web_search calls to gather prospect signals before drafting; if you have no # web search tool, supply the research signals yourself in the prompt or it will drop to Tier 3. # Optionally also copy positioning-basics/ from the same repo — the body reads its output if present. # SKILL-OC.md in the same folder is a token-trimmed variant for OpenClaw; harmless to leave in place.
Kommandon och exempelprompter
/cold-outreach-sequenceBuilds a 5-touch LinkedIn + email cold outreach sequence with personalization tiers
Skills triggas av vanliga förfrågningar — inga kommandon att memorera. Efter installationen aktiverar prompter som dessa skillen (på engelska):
Build a cold email sequence for these prospectsWrite a LinkedIn connection request and follow-upsWarm up this list of cold leads