An agent skill for building UTM-tagged campaign URLs interactively. Paste a URL with a little context ("for the newsletter", "LinkedIn post") and the agent guesses utm_source, utm_medium, and utm_campaign, presents the guesses in a table for correction, and only emits the final URL after you approve or adjust. It preserves existing query strings and fragments, keeps values lowercase and hyphenated, and omits utm_term/utm_content unless they genuinely apply.
Built by 84EM for use with Claude Code and other agent harnesses that load Agent Skills.
| Path | What it is | License |
|---|---|---|
SKILL.md |
The skill: interactive workflow, per-param guess heuristics, formatting conventions, first-pass reply shape | MIT |
Claude Code (personal skill): clone and symlink into your skills directory.
git clone https://github.com/84emllc/claude-utm-builder-skill.git
ln -sfn "$(pwd)/claude-utm-builder-skill" ~/.claude/skills/utm-builderThe skill is then discoverable by its name and description frontmatter. No build step is required.
Paste a URL and say what it is for:
utm link. https://example.com/blog/some-post/ going in my email newsletter
The agent replies with a guess table (param, guess, reasoning) and the assembled URL, then stops for corrections. Confirm or correct, and it emits the final URL in a fenced code block. The one hard rule: no final URL without the correction pause — the review step is the point of the skill.