Menu
SEO & GEO content flywheel โ research a topic, plan it, draft it, de-AI it, run a GEO/SEO + CORE-EEAT quality gate, fan out to social, publish, and feed analytics back into the next cycle. Zero-dependency Markdown skill for any AI agent.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | content-flywheel |
| description | SEO & GEO content flywheel โ research a topic, plan it, draft it, de-AI it, run a GEO/SEO + CORE-EEAT quality gate, fan out to social, publish, and feed analytics back into the next cycle. Zero-dependency Markdown skill for any AI agent. |
| license | Apache-2.0 |
A repeatable, agent-driven content engine. One run produces a complete topic cluster: a primary (pillar) article, supporting cluster articles, platform social copy, and scheduled publishing โ then analytics feed the next cycle.
| Command | Does |
|---|---|
init <brand> | Create a brand profile (voice, audience, authors, URL structure) from schemas/brand.example.yaml. |
run <brand> [--topic "โฆ"] | Execute the full flywheel for one or more topics. |
feedback <brand> | Analyze published performance and tune the plan. |
status <brand> | Report trust level, published count, pillar weights. |
Run phases in order. Each phase is best handled by the agent role noted โ map these to whatever models/tools you have (a strong reasoning model for planning, a fast model for drafting, a research/grounding model for sources, a review model for the gate).
outline.md: H2/H3 framework, thesis, key points, internal-link plan, external-link candidates, CTA, frontmatter + structured-data plan. Inject frameworks/seo-writing-checklist.md + frameworks/geo-techniques.md here. (strongest reasoning model โ plans, does not write prose)frameworks/core-eeat-dimensions.md + GEO rules; list gaps; fix before continuing. (review model)A brand's trust_score.current (0โ1) sets how much runs without a human:
| Trust | Level | Behavior |
|---|---|---|
| < 0.5 | 0 โ manual | Pause at topic selection and at draft review. |
| 0.5โ0.7 | 1 โ semi-auto | Pause at topic selection only. |
0.7โauto_publish_at | 2 โ trusted | Auto-select topics; pause only on low-confidence drafts. |
โฅ auto_publish_at | 3 โ full auto | No pauses. |
Update trust after each human checkpoint: approved as-is +0.02; minor edits +0.01; major rewrite โ0.05.
weekly-plan.yaml (see schemas/weekly-plan.example.yaml).Make runs idempotent: record a per-day ledger and exit cleanly if today is already done; resume partial runs from the last completed phase. After updating any rule, re-run the affected phases โ never reuse stale prior output.
| Platform | Links | Hashtags | Char budget |
|---|---|---|---|
| 0 (use bio link) | โค30 | 2200 | |
| Threads | 0 | โค10 | 500 |
| โค1 | โค5 | ~500 (soft) | |
| X | โค1 | โค3 | 280 |
| โค1 inline | โค5 | 3000 |
A social post without a backing article does not count as a flywheel deliverable.
1 run = 1 published topic cluster = pillar + cluster articles + platform copy + scheduled posts. Articles must pass the CORE-EEAT gate. Accumulate a batch of published items, then run Feedback to close the loop.
schemas/brand.example.yaml โ brand profile template (voice, audience, authors for E-E-A-T, URL rules, trust score).schemas/content-matrix.example.yaml โ content pillars + cluster topics.schemas/weekly-plan.example.yaml โ a 7-day plan.frameworks/ โ GEO, SEO, and CORE-EEAT references injected during planning and review.CORE-EEAT / CITE frameworks adapted from aaron-he-zhu/seo-geo-claude-skills (Apache-2.0). Apache-2.0.