Your cold outreach,
in one Claude Code plugin.
Not another cold-email skill. A full suite: ICP builder, lead sourcing, drafter, sender, reply triage, reporter. All wired together. Reads a portable cold.md file. Runs on FoxReach infrastructure.

Six steps. One command each.
Every step reads and writes files your team can review. No black box, no vendor lock-in. Swap FoxReach for any backend that implements the contract - the skills keep working.
01/cold icp https://your-site.com02/cold leads --count 10003/cold draft Jane, CEO at Acme04/cold send ./leads.csv05/cold triage06/cold report weeklyEach skill owns one file.
cold-icp
Build your ICP from a URL (scrape homepage, case studies) or interactive Q&A. Writes icp.md.
cold-leads
Source leads matching icp.md. Enriches, scores, dedupes, writes leads.csv.
cold-draft
Draft openers, bumps, breakups from cold.md. Refuses banned phrases, cites only from ## Proof.
cold-send
Queue a FoxReach campaign with pre-flight safety: lint, audience check, warmup, sending-limit math.
cold-triage
Sort replies into interested / not-now / OOO / bounced. Drafts responses against ## Objections.
cold-report
Daily or weekly digest: deliverability, replies, booked calls. Runs on cron. Posts to Slack or terminal.
Every skill reads cold.md or icp.md from your repo. Files travel with your team, not with your vendor.
One command. Six skills.
One loop.
$ curl -fsSL https://cold.md/install | bashFOXREACH_API_KEY. Free tier signup →curl -fsSL https://cold.md/install | bash -s -- --skills-onlyThe portable brief every skill reads.
One markdown file in your repo. Eight sections. Any conforming agent reads it before generating a single email.
- →Banned phrases are a hard filter, not a guideline.
- →Voice rules are enforced verbatim - not "inspired by."
- →Only facts from
## Proofcan be cited. No fabricated numbers. - →Audience disqualifiers are respected - won't contact leads who fail your ICP.
# My Product - cold.md ## Identity - who's sending## Audience - ICP + disqualifiers## Value - the offer, in one sentence## Voice - dos and don'ts, obeyed verbatim## Proof - facts the agent may cite## Sequence - opener, bump, breakup slots## Objections - preferred replies to pushbacks## Banned - phrases that never appearOther Claude Code cold-email skills draft. That's one step in a six-step loop.
- - Writes copy when asked
- - No ICP, no lead sourcing, no sending
- - No reply triage, no reporting
- - Every project starts from scratch
- - Config lives in the prompt, not the repo
- → Every step of the loop has its own skill
- → ICP, voice, proof, sequences committed to your repo
- →
/cold reportruns on cron - you never open a dashboard - → Swap sending infra without rewriting anything
- → Works for you, your team, your agents