# Content Skill

> Run one evidence-aware e-commerce content workflow that researches product opportunities, analyzes customer reviews and VOC, optimizes product listings and detail pages, and plans product-image or short-video creative. Use for product research, ASIN or competitor analysis, review CSV/XLSX analysis, negative-review diagnosis, titles, bullets, descriptions, FAQs, Amazon A+ content, Shopee/Shopify/AliExpress listings, product image copy and prompts, storyboards, ad hooks, or an end-to-end research-to-publish content package.

- Skill: `rolandogavino-spec/content-skill` (Agent Skill, multi-file: 27 files)
- Install (CLI): `npx skillmds@latest add rolandogavino-spec/content-skill`
- Raw SKILL.md: https://api.skillmd.com/api/skills/rolandogavino-spec/content-skill/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Data & Analytics
- Author: rolandogavino-spec (https://skillmd.com/u/rolandogavino-spec)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/rolandogavino-spec/content-skill

---


# Content Skill

Coordinate three imported Orkas agent roles and four imported merchandising skills as one Codex workflow. Preserve their capability boundaries while passing evidence and decisions forward between stages.

Source: Orkas-AI/Orkas-Awesome-AgentSkills, e-commerce collection. The bundled source remains under its original MIT license in `references/ORKAS-LICENSE`.

## Route the request

Choose the smallest useful path:

- Product/category viability, ASINs, competitors, price bands, profit assumptions, sourcing risk, or trend validation: use **MerchResearcher** with `merch-research`.
- Uploaded/exported reviews, ratings, VOC, negative-review causes, competitor feedback, or reputation analysis: use **MerchReviewer** with `merch-review`.
- Titles, bullets, descriptions, keywords, FAQs, Amazon A+, detail pages, image copy, visual prompts, hooks, or storyboards: use **MerchPageOptimizer** with `merch-listing` and/or `merch-creative`.
- End-to-end content work: run Research -> Review when feedback exists -> Page Optimization -> Creative. Skip any stage that has no relevant input.

For mixed requests, keep one shared evidence ledger so downstream stages can distinguish verified facts, user-provided facts, assumptions, inferences, and missing evidence.

## Load the role and skill instructions

Before performing a stage, read the corresponding agent definition and relevant imported skill completely. Then read only the skill references it routes to.

| Stage | Agent definition | Skill entry point |
|---|---|---|
| Research | `references/agents/MerchResearcher.json` | `references/skills/merch-research/SKILL.md` |
| Review | `references/agents/MerchReviewer.json` | `references/skills/merch-review/SKILL.md` |
| Listing | `references/agents/MerchPageOptimizer.json` | `references/skills/merch-listing/SKILL.md` |
| Creative | `references/agents/MerchPageOptimizer.json` | `references/skills/merch-creative/SKILL.md` |

Treat Orkas-specific skill IDs, tool names, runtime settings, and routing metadata as source context rather than Codex configuration. Use currently available Codex tools and applicable local skills for spreadsheets, documents, web research, or image generation.

## Run the Content Skill workflow

1. Frame the job. Identify product, category, platform, country/market, customer, SKU/price, brand voice, goal, supplied files/links, competitor scope, deadline, and desired outputs. Ask only for information that materially changes the result; otherwise state conservative assumptions.
2. Establish evidence. Prefer user-provided product facts, legal exports, authorized analytics, supplied competitor materials, and public pages. Browse when the user asks for current external research or the answer depends on changing facts. Never bypass logins, CAPTCHA, anti-bot controls, paid limits, or platform terms.
3. Research when needed. Evaluate demand signals, competition, price band, differentiation, economics assumptions, supply/fulfillment risk, compliance, confidence, and next validation steps. Do not equate social buzz, BSR, displayed sales, or third-party estimates with confirmed demand.
4. Review feedback when available. Remove or suppress personal identifiers, document the sample and denominator, classify VOC and root causes, preserve short anonymized evidence snippets, flag sample bias, and rank actions by impact, confidence, cost, reversibility, and urgency.
5. Optimize the listing. Convert verified benefits and objections into platform-appropriate titles, keywords, bullets, descriptions, FAQs, A+ or detail-page structure, and image sequence. Mark unsupported claims as needing evidence or rewrite them conservatively.
6. Plan creative. Translate verified selling points and VOC into main-image copy, detail-page modules, product-image prompts, video hooks, storyboards, CTAs, and A/B tests. Do not generate media unless the user asks for generation; then use the applicable image/video capability and its safety rules.
7. Review the package. Check platform fit, unit/language consistency, privacy, claims, trademarks, medical or regulated-category risk, unsupported comparisons, misleading scarcity, unauthorized assets, and internal contradictions.
8. Deliver one coherent package. Include assumptions, source/evidence summary, stage outputs, missing evidence, compliance/manual-review items, prioritized actions, and next tests. Keep intermediate role handoffs internal unless the user asks to see them.

## Handoff contract

Pass these fields between stages when present:

- Product facts and source labels
- Target platform, market, audience, price, and positioning
- Competitors and comparison limits
- Buyer motivations, objections, positive drivers, pain clusters, and root causes
- Approved claims, claims needing evidence, and prohibited claims
- Keywords, page opportunities, creative angles, and proof requirements
- Confidence, data limitations, and unresolved decisions

Never invent certifications, rankings, sales, reviews, endorsements, discounts, product performance, or competitor facts. For food, supplements, cosmetics, children's products, medical devices, pet food, and electronics, use conservative language and explicitly request human compliance review.