← ClaudeAtlas

long-runnerslisted

Find the Meta ads competitors keep paying to run for months, and explain why they might hold up. Use for "what's evergreen in my category", "whose ads last longest", "what keeps working", "what should we model", or before planning a new batch of creative.
seercraft-io/seer-skills · ★ 0 · AI & Automation · score 72
Install: claude install-skill seercraft-io/seer-skills
# Long-runners Brands switch off ads that don't pay. The ones still running after 60, 120, 300 days are the closest thing to proof the Ad Library gives you. Done means: `seer/reports/long-runners-<date>.md` with the top long-running ideas across rivals, what they share, and what to borrow. Read `${CLAUDE_SKILL_DIR}/../seer-start/reference/house-rules.md` first. `$S` = `${CLAUDE_SKILL_DIR}/../seer-start/scripts`. ## Steps 1. Make sure every rival in `seer/rivals.md` has a read from the last 7 days. 2. `python3 $S/ideas.py seer/ads/<all rivals>-<date>.json --top 12` 3. From the "still running longest" tables, take the top 10 ideas across brands, at most 3 per brand. Skip catalog/dynamic ideas (no hand-written copy to learn from) unless that's all there is. 4. For each: hook (first line, verbatim), angle, proof, offer, format, versions. Versions matter: an old idea with many versions is being actively scaled, not just forgotten. 5. Look for what they share: a hook type, an angle, a format, a promise. Only call it a pattern if 3+ of the 10 share it ("4 of 10"). 6. **Depth check.** Impression-sorted reads lean to recent ads. If a brand's longest idea in your read is under 60 days but it has hundreds of live ads, say its real long-runners are probably deeper than you read, and offer a deeper read. ## Output - Table: days running · brand · idea (cited) · versions · format · hook (verbatim, short). - For each of the top 5: one line on why it might hold up (your r