# How Elad Hefetz built a Claude skill that pitches podcast hosts and got a 70% reply rate

The author of this Airfleet post wanted a repeatable way to get booked on B2B podcasts without paying a PR agency. He built a Claude skill on top of Gal Ko's PodStar 5R framework. The skill finds shows through the Apple iTunes Search API and pulls host emails and recent episode titles from RSS feeds. It uses Bright Data to find host LinkedIn profiles, then scores each show and writes a personalized pitch. Lemlist sends a five-step sequence that stops on reply. He reports a reply rate of almost 70 percent and 7 guest offers.

| | |
|---|---|
| **Who** | Elad Hefetz, per the brief (the source text does not show his name or title), writing on the Airfleet blog ([airfleet.co](https://www.airfleet.co/blog/)). X: [UNVERIFIED] not checked |
| **Problem** | Getting booked as a podcast guest consistently, "without a PR agency charging thousands for it" |
| **Stack** | Claude skills, parallel research agents, Apple iTunes Search API, RSS feed parsing, Bright Data, Lemlist, Python scripts, a dedicated sending domain |
| **Confidence** | Medium (first-person, but author named in the page byline metadata; number of hosts pitched not stated) |
| **Last verified** | 2026-10-16 |

## The problem
The author describes himself as an introvert. He says, "Reaching out to strangers in real life is hard for me." He wanted to "stop waiting to be discovered and build a repeatable machine to get booked." He set out to solve the part that comes before the mic: "consistently getting booked without a PR agency charging thousands for it."

## The blueprint (as he describes it)
1. **Use a framework the agent can follow.** He built around Gal Ko's PodStar 5R system: Reframe, Reveal, Research, Reach, Resonate. He automated the three steps he says scale: Reveal, Research and Reach.
2. **Discover shows, not people.** His first instinct was to reuse his LinkedIn event-prospecting pipeline. He calls that a mistake, because "Podcasts have no public attendee roster." He switched to the Apple iTunes Search API, which is keyless and returns each show's RSS feed URL. He ran it across dozens of search terms and deduped the results by feed URL. The first sweep surfaced around 600 shows.
3. **Enrich from RSS.** Each feed exposes the owner email, website, language, last episode date and last five episode titles. He calls the recent titles "the single best signal for personalizing a pitch." About 65 to 70 percent of shows expose a direct owner email this way.
4. **Find the host on LinkedIn with Bright Data.** RSS gives nothing for LinkedIn. He used Bright Data to resolve and verify host profiles at scale, so each high-fit show had a second channel alongside email.
5. **Score and match an angle.** Each show is scored on fit, reachability, recency and audience size, weighted toward fit. Shows dormant beyond 90 days are dropped. Off-topic, non-English and thin shows are cut automatically. Each remaining show is matched to the single best angle from his talking-point library. For example, an SEO show maps to GEO.
6. **Personalize with fixed rules.** The main rule is "never pitch Airfleet." Airfleet gets one short credibility clause at most, and the rest of the message is about value to the audience. The skill paraphrases a recent episode rather than copying the title. It keeps emails to around 90 to 130 words and LinkedIn notes to 50 to 70. It uses the host's real first name. Each show gets a subject, an email body, a shorter LinkedIn DM and a follow-up with a fresh angle. He fans this work out across parallel research agents.
7. **Hand sending to Lemlist.** Each show becomes a Lemlist lead, and all the bespoke copy sits in custom variables that the campaign template references. The sequence runs on days 0, 2, 5, 9 and 14:
   - Day 0: personalized email.
   - Day 2: profile visit and a short connection note.
   - Day 5: follow-up in the same thread.
   - Day 9: a message if connected and no reply.
   - Day 14: a final short nudge, then stop.

   Lemlist stops the sequence when someone replies. He sends from a dedicated sending domain rather than the primary domain.
8. **Approve before sending.** In the FAQ he writes that "Every draft is previewed before anything sends". He approves the batch, and Lemlist handles sequencing and follow-ups.

## Stated results
- Reply rate of "almost 70 percent" (the headline rounds this to 70%).
- 7 guest offers.
- Pipeline: 600 shows discovered, 109 ranked, 81 high-fit and 66 with a direct host email, all active within the last 90 days.
- "Several yeses came on the follow-up, not the first email."
- He says the replies converted to recordings, but no recording count is stated.
- The number of hosts actually pitched is not stated, so the reply-rate denominator is unknown.

## Caveats
- This is a post on Airfleet's company blog, which doubles as marketing for Airfleet. The supplied source text does not show the author's name. The attribution to Elad Hefetz comes from the brief and needs checking against the live page.
- The post offers a downloadable kit with the scripts, scoring model, personalization rules and Lemlist blueprint. The source does not say whether it is free or paid.
- The 1 to 5 percent "usual" cold reply rate is the author's claim with no source given. The listener statistics come from third parties (one is attributed to Kitcaster), not from this campaign.
- The sequence table in the source has an empty channel column. The channel for each step is implied by its content but not stated.
- The workflow leans on Gal Ko's PodStar framework. The post links to the PodStar community on Skool. It names Bright Data and Lemlist but does not say whether either is a partner.
- He says the setup needs "a few Python scripts" and a Lemlist connection. He also calls 70 percent "the ceiling, not the baseline."
- X: [UNVERIFIED] not checked.

## Sources
- https://www.airfleet.co/blog/podcast-guesting-outreach-skill/ (primary: first-person build write-up on the Airfleet blog)

---
*Want agents like these working your distribution? [See the Agent Stack Teardown](/agent-stack-teardown). Get new teardowns weekly: [subscribe](/subscribe) (double opt-in).*