What do I need to use Promoter Scan?

Where the transcripts are.; The minimum count worth reporting; Anything that must never appear in the report.

Is Promoter Scan free to use?

Yes — run it free in the browser playground (no API key needed for the sponsored trial, or bring any provider key), or install it into Claude Code, Cursor, ChatGPT and more via npx pm-claude-skills. MIT-licensed and open source.

Promoter Scan

Q: How do I know the Promoter Scan output is good?

The skill verifies itself against its own quality checks: Every example prompt has been through redaction and contains no email, key, number sequence or URL; Each pattern is one job; families ("summarise X" for five different Xs) are flagged as maybe with a note on how to split; The score components are shown, not just the total; One-off sprees (ten asks in one afternoon, never again) are called out rather than promoted.

🔵 Stable 🔷 SkillSpec L3 ⚙ ships an executable helper pm-skill-promoter

Find the requests you keep making to your AI assistant by scanning Claude Code transcripts and claude.ai exports, clustering them by intent, and ranking each recurring pattern by how promotable it is into a reusable skill. Use when asked what do I keep asking for, scan my transcripts, find repeated prompts, which of my prompts should become skills, or to start the promote loop. Produces a ranked report of recurring patterns with frequency, redacted example prompts, the variation between them, and a promotability score.

▶ Run it free — no key needed 📝 Grade your existing draft View SKILL.md ↗

🗣 Say this to your agent

“What do I keep asking Claude for?”“Scan my transcripts and tell me which prompts should be skills.”“Find my repeated prompts from the last month.”“Start the promote loop on my Claude Code sessions.”

What to give it

▸Where the transcripts are. — Claude Code keeps session transcripts as `*.jsonl` under `~/.claude/projects/<project>/`. A claude.ai export is a `conversations.json` inside the downloaded archive. Either or both work; a folder is fine.

▸The minimum count worth reporting — (default 3) and, if the user wants, a date range to limit the scan

▸Anything that must never appear in the report — beyond the built-in redaction (client names, project codenames)

✅ The bar it holds itself to

Every skill in this library self-verifies — these are this skill's own quality checks, straight from its definition.

✓Every example prompt has been through redaction and contains no email, key, number sequence or URL

✓Each pattern is one job; families ("summarise X" for five different Xs) are flagged as maybe with a note on how to split

✓The score components are shown, not just the total

✓One-off sprees (ten asks in one afternoon, never again) are called out rather than promoted

✓A suggested skill name is given for every promote verdict, checked against the library for collisions

⚠️ What it refuses to do

**Quoting a prompt verbatim because it looked harmless.** Redact first, judge later.

**Promoting by frequency alone.** Fifty variations of "help" are not a skill.

**Merging jobs that share a verb.** "Summarise a thread" and "summarise a paper" want different frameworks.

**Scanning other people's transcripts.** The loop is for the user's own usage, with their consent to read it.

Install

npx pm-claude-skills add --agent claude   # or codex · cursor · gemini · hermes
# or one-line MCP (every skill, any client):
claude mcp add pm-skills -- npx -y pm-claude-skills-mcp

Related skills

🔌 Embed this skill

Drop this on your blog, docs, or site — it renders a "Run this skill" card:

<div data-pm-skill="promoter-scan"></div>
<script src="https://mohitagw15856.github.io/pm-claude-skills/embed.js" async></script>

💬 Discussion

Promoter Scan is one of 1182 open-source professional AI agent skills — all SkillSpec L3. Try them all in the browser · ⭐ Star on GitHub · Browse the full catalog