From 6d4847c1ed2bff709724fd712429766eab619cbb Mon Sep 17 00:00:00 2001 From: Ty Date: Thu, 6 Aug 2026 00:47:23 +0000 Subject: [PATCH] Add parasite-seo skill for Leonard / Digital Operations Partner buy-intent lead engine --- skills/parasite-seo/SKILL.md | 126 +++++++++++++++++++++++++++++++++++ 1 file changed, 126 insertions(+) create mode 100644 skills/parasite-seo/SKILL.md diff --git a/skills/parasite-seo/SKILL.md b/skills/parasite-seo/SKILL.md new file mode 100644 index 0000000..ff1895f --- /dev/null +++ b/skills/parasite-seo/SKILL.md @@ -0,0 +1,126 @@ +--- +name: parasite-seo +description: Execute or manage buy-intent parasite SEO for young domains or client lead generation. Use when ranking high-purchase-intent long-tail queries via Medium Substack YouTube GitHub Instagram, automating programmatic SEO for ICP + location combinations, or productizing this as a Digital Operations Partner service for SMBs. Triggers on parasite SEO, buy-intent keywords, GEO optimization, local service lead gen automation, or client content pipelines that skip weak-domain ranking fights. +--- + +# Parasite SEO (Buy-Intent) + +## Overview + +Skip the multi-year ranking war on weak/young domains. Generate and publish thousands of ultra-specific, high-purchase-intent long-tail answers on platforms Google and AI answer engines already trust. Automate the pipeline so it runs nightly. This is a core lead-generation capability for Digital Operations Partner clients (local services, professional practices, productized SMBs). + +## When to Activate + +- Client or internal domain is <3 years old or has low authority +- Goal is buyer acquisition (not vanity traffic) +- Queries follow the pattern `best/how much/ alternative for in ` +- Need scalable, automated content that also feeds ChatGPT / Perplexity citations +- Building or operating a managed "Parasite SEO as a Service" package inside Digital Operations Partner + +## Core Principles + +1. Target only buy-intent long-tails. Volume is secondary to commercial intent. +2. Publish on high-authority parasite domains (Medium, Substack, YouTube, GitHub, Instagram). Rent their trust. +3. Every article must contain a unique statistic drawn from real client/internal data. Generic advice is never cited. +4. Track every keyword with a unique constraint so crashes cannot double-post. +5. Verify the public URL is live before marking success. Automation often reports false positives on drafts. +6. Failed keywords return to the pool. Never permanently consume a keyword on a failed publish. + +## Required Inputs (collect or confirm before running) + +- Parent keywords (10–30 trunk terms, e.g. "best payroll software", "gusto alternative") +- ICP / vertical list (e.g. personal injury lawyers, med spas, dental practices, yoga studios — prioritize by customer LTV) +- Location list (cities, metros, or "near me" variants) +- Real question list from Google Search Console (or equivalent) filtered to questions +- Platform accounts + credentials or API access (Medium, Substack, YouTube, GitHub, Instagram) +- Access to proprietary data sources for unique stats (billing, analytics, support tickets, reviews) +- Target publish cadence (default: 1 original article per platform account per night) + +## Keyword Universe Generation + +1. Start with parent keywords. +2. Cross with every ICP and every location: + `best for in ` + `how much does cost for in ` + ` alternative for ` +3. Score and sort so highest commercial-value combinations (high LTV verticals first) publish earliest. +4. Store in a durable table with unique constraint on the full keyword string + platform. + +Example scale: 31 parents × 86 verticals × 99 cities = 166k+ unique queries. Generate once, publish for years. + +## Content Format (mandatory for both SEO and GEO) + +Every piece must contain: + +- Quotable summary block (3–5 sentences) at the absolute top — this is what AI engines lift. +- Direct answer to the query in the first paragraph (no fluff intro). +- FAQ section built from the real Search Console questions collected earlier. +- Exactly one image — filename and alt text = the target keyword. +- Internal links to earlier posts published on the same platform account. +- At least one unique, non-invented statistic from real data (billing history, product analytics, support tickets, review aggregates, etc.). + +Never invent numbers. If proprietary data is unavailable, stop and request it. + +## Automation Prompt (hand to Claude / Hermes / Leonard) + +Use this (or a refined version) as the core routine: + +``` +Build me a nightly parasite SEO routine. Generate a keyword universe from my parent keywords crossed with buyer verticals and cities, scored so the most commercial combinations publish first. Each night, publish one original article per account across my platform accounts, one keyword each, never repeating a keyword. Track what published in its own table with a unique constraint on the keyword so a crash cannot double post. Every article gets a quotable summary block up top, an FAQ from my real Search Console questions, one generated image with the keyword as filename and alt text, and internal links to my earlier posts on that same account. Only cite statistics from my own database, never invented ones. Record what failed each night and feed it into the next run. +``` + +After the first run, instruct the model to improve itself on every failure. + +## Platform Execution Notes + +- Prefer real APIs where available. +- For platforms without APIs, use browser automation (Claude browser use or equivalent Playwright/Puppeteer). +- After every publish, verify the returned URL is publicly accessible and not a draft/editor page. +- Record the verified public URL, not the editor URL. + +## Failure Handling (critical) + +- If publish reports success but URL is not public → treat as failure, return keyword to pool, log the exact failure reason. +- Login / auth failures → do not consume the keyword. +- Image generation or data lookup failures → abort that article, return keyword, continue with next. + +## Integration with Digital Operations Partner + +This skill is a productizable module inside the Digital Operations Partner offering: + +- Position as "Managed Buy-Intent Lead Engine" or "Parasite SEO Pipeline" for clients whose own sites cannot rank. +- Typical clients: local service businesses, professional practices, and young SaaS/product companies (examples already in use: salons, medspas, law firms). +- Delivery model: one-time setup of keyword universe + platform accounts + data connectors, then monthly recurring management + reporting of published pieces and resulting leads/citations. +- Leonard (or Hermes agents) can own the nightly execution, verification, and weekly performance summary for each client. +- Unique stats come from the client's own systems (or agency-collected review/performance data) so every piece has proprietary signal that competitors cannot copy. + +When operating for a client, always: + +1. Confirm data-sharing agreement for the unique statistics. +2. Use client-branded or neutral voice as specified in the engagement. +3. Report verified public URLs and any ranking/citation signals back into the client dashboard or weekly ops report. + +## Tracking Requirements + +Maintain a single source of truth table (or spreadsheet / DB) with at minimum: + +- keyword (unique) +- platform +- status (queued / published / failed) +- published_url (only when verified public) +- published_at +- failure_reason (if any) +- unique_stat_used +- next_retry_after (for failed items) + +## Safety and Quality Rules + +- Never invent statistics or fabricate data sources. +- Never mark a keyword consumed on a failed or unverified publish. +- Prefer quality and uniqueness over volume on any single night. +- If the client has an existing brand voice or compliance requirements, load those first and enforce them. + +## Related Resources + +- Original playbook source: the Aug 2026 @borjafat thread on buy-intent parasite SEO. +- For platform-specific publishing scripts or data connectors, place them under `references/` or `scripts/` and load on demand.