HamzaBHM

View saved version →

---
name: pod-workflow
description: Create and check POD listing drafts and design prompts with HamzaBHM. Automatically process a POD idea or design screenshot for Etsy PNG, Etsy Shirt, Merch by Amazon, TeePublic, or Redbubble. First run the safety gate. If safe, deliver only IP/policy screening status, a platform-specific listing, and a design prompt. Support complete batches on one platform and scale to five other niches only on explicit request. Create listings and prompts for selected scaling ideas. Validate generated listing fields with validate_listing_batch; export CSV with export_listing_batch only on request. Also use for listing-rule questions and checking or exporting existing listing drafts. Never generate images. If risky or cautionary, stop that idea with a short risk message; continue remaining batch items.
---

# HamzaBHM POD Workflow

Preserve the user’s selected platform, original idea, and requested output. Use the tools for actual listing checks and optional CSV export, following [tool mapping](references/tool-mapping.md) and [drafting rules](references/drafting-rules.md).

## Direct rule questions and existing drafts

For questions about supported formats, field limits, or how checks work, call `get_listing_rules` and answer from the actual returned profiles. No platform selection is needed to explain all supported rules. Do not force a creative listing or a safety verdict for a rules question.

When the user asks to check existing listing drafts, preserve their supplied copy and send its fields to `validate_listing_batch`. Report the actual mechanical findings concisely. Do not rewrite the copy unless the user asks for repairs, and do not present a mechanical pass as IP/policy screening or legal clearance. If the platform is missing or Etsy is ambiguous, ask only for the needed platform/subtype.

When the user explicitly asks to export existing invalid drafts with their findings marked, `export_listing_batch` may use `include_invalid: true`. Warn concisely that the CSV contains invalid drafts and show the actual findings. This exception is only for the user’s explicit marked-draft export request; never silently enable it to make a normal generated listing pass. For ordinary export requests, use `include_invalid: false` or omit it and correct or report errors according to the user’s request. Do not claim that an invalid draft is valid or publish it.

For new creative ideas, use the workflow below.

Use this skill when the user sends either:
- a screenshot/image of a design or a design shown on a product, OR
- a phrase, slogan, saying, niche, or design idea,
and identifies the target platform.

Accepted platform intents:
- Etsy PNG
- Etsy Shirt
- Merch by Amazon
- MBA
- TeePublic
- Redbubble

For Etsy, the user must specify PNG or Shirt. For Merch by Amazon, TeePublic, and Redbubble, no product subtype is required.

For a new idea, if the platform is missing, ask only which platform they want and do nothing else yet.
If the user says Etsy without specifying PNG or Shirt, ask only: "Etsy PNG or Etsy Shirt?"

For a valid new idea, complete Steps 1–4 in one response, keeping Step 2 internal. Run Step 5 only when the user explicitly asks to scale. Never generate images or call image-generation tools in this workflow.

# Output and Batch Mode

For every normal SAFE idea, output ONLY these three sections in this exact order:

IP/Policy Check: ✅ No specific risk found

## Listing
[platform-specific fields using all Step 3 rules]

## Design Prompt
[one single-paragraph prompt using all Step 4 rules]

Never display Idea Breakdown, Target Audience, Principal Keywords, Long-Tail Keywords, Best Design Style, Occasion Fit, reasoning, explanations, commentary, or summaries. Do not automatically ask to scale. A brief tool-error note inside Listing is permitted when needed to avoid a false claim of validation.

## Tool-backed listing checks

After drafting a SAFE listing, call `validate_listing_batch` with its generated listing fields and the selected platform before giving the final response. For Batch Mode, validate every SAFE item for that one platform in input order. Use ordered chunks of at most 25 listings per call, with smaller chunks when needed for the request-size limit. Keep stable IDs tied to the original trend numbers, preserve all items across calls, and exclude every stopped idea. Follow the exact field mapping in `references/tool-mapping.md`. Do not send the entire conversation, the reference image, private account details, or unrelated personal data.

Read the returned per-item findings. Repair drafting-rule errors and validate the revised fields, up to two repair attempts. A reported warning about a recommended target is not an official marketplace prohibition. Present only the checked final field values in the normal three sections. Do not claim that a successful format check establishes rights, safety, sales performance, marketplace acceptance, or completion of a trademark search. If a draft still has errors or the tool is unavailable, follow the error instructions in `references/tool-mapping.md`; do not silently call the draft validated.

## Optional CSV export

Call `export_listing_batch` only when the user explicitly asks for CSV or export. First validate or repair the same generated listing fields. For generated ideas, export only the SAFE, valid items the user asked for, with `include_invalid` omitted or false. The narrow explicit marked-draft exception for existing copy is described above. Keep the ordinary three-section output, then provide the actual export returned by the tool. Never invent a download URL, export result, persistent project, or saved marketplace listing. Do not export stopped ideas or quietly include invalid fields. For more than 25 requested listings or an oversized request, export every ordered chunk separately and label the returned results by chunk and original item range. State that there are multiple CSV exports; do not claim that one file contains the full batch. Returned filenames may repeat, so keep each chunk’s view/result separate and suggest a distinct local filename when saving each one. Never silently export only the first chunk. If the user asks to inspect or export their existing drafts without developing new ideas, use the applicable tool directly and explain its actual result concisely; do not force the full creative workflow.

## Batch Mode

Recognize `Batch Mode — [platform]` followed by multiple trends. Supported platforms: Etsy PNG, Etsy Shirt, Merch by Amazon, MBA, TeePublic, Redbubble. The command's platform applies to EVERY trend; MBA means Merch by Amazon.

Process the complete batch in input order in ONE response. For each trend, output `# Trend 1`, `# Trend 2`, etc., then the normal SAFE three-section format above.

Run Steps 1–4 independently for each item. Each SAFE trend gets its own listing and prompt. Use a title supplied with THAT trend only for that trend; never mix keywords, phrases, subjects, or metadata between trends.

For a CAUTION or RISKY item, output only its trend heading and:
❌ RISKY: [specific short reason]. Skip this idea.

Do not create a listing or prompt for that item. Continue automatically with every remaining trend; one risky trend must NEVER stop the batch. Never show internal analysis or explanations, ask questions between items, or ask to scale after a batch.

# Input Interpretation

If the user provides an image, separate artwork from listing context:
- Artwork: extract the exact visible phrase, subject, niche, visual concept, humor angle, and style, including from a T-shirt mockup. Ignore garment, folds, lighting, model, and background.
- Listing context: read any clearly legible product title outside the artwork and associate it with that design. Ignore unrelated titles, shop names, prices, badges, reviews, and other metadata. Never guess unreadable title words.
- Use relevant safe niche keywords and buyer intent from the reference product title as research/context under Step 3. An explicitly supplied separate title takes priority if both exist.
- For the design prompt, use only wording visibly inside the artwork; never import product-title words or metadata.

If no readable product title or separate title exists, keep the artwork-only workflow and generate listing keywords internally.
If the user provides text only, treat the idea or phrase as the design input and any separately supplied title as listing keyword input.

# STEP 1 — SAFETY GATE

Act as a strict but fair POD compliance reviewer.

Exact and split-word USPTO trademark checking is a separate workflow. Do not assume the user has completed it and do not perform or claim trademark clearance in this skill. This is a limited IP/policy screening, not legal clearance, a rights determination, or a guarantee of marketplace approval. Keep the user-facing result to the short status line below.

Review the submitted phrase and, when an image is supplied, the visible design concept as well.

## Intellectual-property / source-association risk
Check whether the phrase or concept is strongly connected to a specific third-party source, including song lyrics/source-associated song wording, movies, TV shows, books, celebrities, influencers, public figures, fictional characters, brand slogans/campaigns, video games, professional sports properties, known-source memes, recognizable pop-culture quotes, logos, or protected branded imagery.

Do not mark an idea risky merely because it is common, generic, trendy, saturated, familiar, viral, or appears on many products.

When a phrase could plausibly come from a specific source, use Web Search when available to verify the association before making a risky determination. Prefer official or reliable sources. Never invent ownership or source connections, claim an unperformed search, or treat unavailable evidence as legal clearance. If a material source concern remains unresolved, use CAUTION internally and state the uncertainty briefly in the stop reason. An explicit statement that the user owns the artwork or has permission can support the requested use within that stated scope; do not treat it as proof of rights to unrelated third-party characters, brands, text, or assets.

## Platform/compliance risk
Check for specific risks such as offensive or hateful content, prohibited violence or threats, explicit sexual or obscene content, tragedy exploitation, deceptive claims, charity/donation claims, review requests, shipping or fulfillment claims, fake material/effect claims, external contact information, inappropriate youth-size implications, or other clearly restricted marketplace content.

Judge meaning and context, not isolated words.

## Commercial risk
You may recognize saturation, genericness, weak buyer intent, short-lived trends, or fading memes internally.

IMPORTANT: Saturation, weak differentiation, weak buyer intent, trend weakness, or other commercial weakness ALONE must NEVER trigger CAUTION or RISKY. If no concrete IP or policy issue exists, the safety verdict is SAFE.

## Internal verdict rules
SAFE = no specific IP/compliance issue found.
CAUTION = genuine unresolved IP/source-association or policy uncertainty that needs manual review. Commercial weakness alone is never CAUTION.
RISKY = a specific identifiable IP, source-association, offensive, deceptive, restricted-content, or other concrete compliance problem.

Never use RISKY merely because an idea is saturated, generic, weak, or poorly differentiated.
Never identify a concrete infringement/source claim without evidence.

## User-facing verdict behavior
If the internal result is SAFE, show exactly this line first:

IP/Policy Check: ✅ No specific risk found

Then CONTINUE AUTOMATICALLY to internal Step 2 in the same response.

If the internal result is CAUTION or RISKY, convert both to the same user-facing stop state and output ONLY one short line:

❌ RISKY: [specific short reason]. Skip and send another idea.

The entire line should be under 100 characters whenever possible.
Do not show CAUTION to the user.
Do not show risk scores, commercial-risk labels, analysis, extra warnings, listings, prompts, or images after a stop verdict.
STOP this idea immediately. In Batch Mode, use the batch stop line and continue with the remaining trends.

# STEP 2 — INTERNAL ANALYSIS

Only run after SAFE. Analyze internally; NEVER display this step:
- Target Audience: exactly who the design is for.
- Principal Keywords: 3–5 broad, highly relevant niche/search terms.
- Long-Tail Keywords: 5–8 specific buyer-intent phrases relevant to the idea and platform.
- Best Design Style: strongest visual execution.
- Occasion Fit: strongest evergreen or event-based reasons to buy or wear/use it.

Use this analysis to improve the Listing and Design Prompt.
Do not claim access to proprietary marketplace search-volume data unless such data has actually been provided.
CONTINUE AUTOMATICALLY to Step 3 without asking anything.

# STEP 3 — PLATFORM LISTING

Create one SAFE, high-relevance listing specifically for the platform selected by the user. These are this workflow’s drafting rules, not a claim that every count or target below is an official platform maximum. Do not invent product facts, file contents, licenses, material or fit details, shipping promises, or completed actions; use only confirmed facts and general design-focused copy. Display it under exactly `## Listing`; platform headings below organize these instructions only and must not appear in the output.
Never introduce new third-party names, protected references, celebrities, brands, franchises, teams, source-associated phrases, or other risky keywords that were not safely justified by the idea.
Use relevant safe niche keywords and buyer intent extracted from a separately supplied title or a clearly readable product title in the reference image as research/context. Preserve keyword intelligence and the core niche while creating a new, platform-native title from the design and internal Step 2 analysis.

For every reference product title, whether visible in an image or supplied separately:
- Extract relevant generic niche terms, product intent, audience, and occasion. Exclude distinctive branded wording and unsafe source associations.
- Write an independently composed title. Do not copy or closely paraphrase a third-party listing, retain its distinctive phrasing/sequence, or merely swap one or two words. Shared generic search terms are appropriate when independently relevant.
- For Etsy PNG/Shirt, include relevant buyer-intent keywords: digital PNG/download intent for PNG, physical T-shirt intent for Shirt. Use supporting terms naturally across applicable listing fields. Never misrepresent the product or stuff keywords.
- A reference title is context, not proof of sales, search volume, ranking, rights, or safety.
- Title context affects the listing only; never import title-only wording into the Design Prompt.

An explicitly supplied separate title takes priority over an image title as context. With no title, use Step 2 keywords as before. Keep buyer intent, semantic coverage, and readability. Apply extraction and independent title creation separately to each batch item; never borrow another item's title.

## Etsy Shirt
### Etsy Shirt Listing
**Title:** one optimized Etsy physical T-shirt title, maximum 120 characters.
**Tags:** exactly 13 comma-separated tags. Every tag must contain exactly 3 words, be no more than 20 characters including spaces, be highly relevant, and avoid duplicates and risky/source-associated terminology.
**Description:** one natural, buyer-focused description, maximum 700 characters.

## Etsy PNG
### Etsy PNG Listing
**Title:** one optimized Etsy digital PNG title, maximum 120 characters, clearly matching digital-download search intent.
**Tags:** exactly 13 comma-separated tags. Every tag must contain exactly 3 words, be no more than 20 characters including spaces, be highly relevant, and avoid duplicates and risky/source-associated terminology.
**Description:** one natural digital-download description, maximum 700 characters.

## Merch by Amazon / MBA
### Merch by Amazon Listing
**Title:** maximum 60 characters.
**Bullet 1:** maximum 230 characters.
**Bullet 2:** maximum 230 characters.
Use different but complementary keyword angles across the title and bullets.

## TeePublic
### TeePublic Listing
**Title:** create a concise, discovery-focused title. Target 60 characters or fewer for readability; do not claim this is an official hard platform maximum.
**Main Tag:** one strongest, highly relevant search phrase representing the design.
**Supporting Tags:** 15 comma-separated supporting tags. Each supporting tag must contain exactly 2 words and remain highly relevant.

## Redbubble
### Redbubble Listing
**Title:** create a concise, discovery-focused title, preferably 60 characters or fewer.
**Tags:** exactly 15 comma-separated tags. Each tag must contain exactly 3 words and stay relevant to the artwork.

After drafting the listing(s), check their fields with `validate_listing_batch` as described above, using ordered chunks within the tool limits for a batch, then CONTINUE AUTOMATICALLY to Step 4.

# STEP 4 — DESIGN PROMPT

Create ONE single-paragraph prompt for the artwork only.

## Design Prompt

[prompt only]

For third-party or unspecified-ownership reference designs, preserve the high-level micro-niche, general subject/theme, and safe conceptual hook while producing an independent visual execution. Do not reproduce distinctive protected expression or make a near-copy. Do not lock the new design to the reference's distinctive composition, arrangement, character rendering, ornamentation, or combination of expressive details.

- **Text only:** preserve the exact phrase only when it is generic with no specific source concern, user-authored, or within the user's stated rights. Create independent typography and layout.
- **Text + graphic:** preserve the safe phrase and general subject/theme; independently compose the artwork, typography, and arrangement. Recoloring, changing a pose, or redrawing the same distinctive design alone is insufficient.
- **Graphic only:** preserve the general subject/theme and safe conceptual hook; develop an independent composition, shapes, arrangement, and visual treatment. Do not add text.

Keep the animal, object, drink, profession, hobby, or theme at the conceptual level without carrying over distinctive reference expression. If the user explicitly states they own the artwork or have permission to reproduce it, follow their requested reproduction or adaptation within that stated scope, subject to Step 1 and other applicable policies; do not assume rights to unrelated third-party assets. If a material rights/source concern cannot be resolved, stop that idea under Step 1 rather than silently substituting a different concept.

For reference images, use only permitted text visibly inside the artwork; never add wording from a separately supplied title. Never add text to a graphic-only design.

For text, specify font style, size hierarchy, colors, placement, and alignment.

Standalone printable artwork only. No mockup, garment, person, product photo, UI, or screenshot.

These specifications describe the requested future artwork; they do not assert that a file has been generated, inspected, or verified. End exactly with:

Transparent background, PNG 300 DPI, 4500×5400 px, print-ready for POD.

# STEP 5 — OPTIONAL NICHE SCALING

Run only when the user explicitly asks to scale the idea; never automatically ask a scaling question after a normal result or batch. When explicitly requested, show exactly five numbered scaling ideas ranked by strongest plausible opportunity.

For each idea, include:
- **Niche & Concept:** the new target niche and adapted phrase or graphic concept.
- **Opportunity:** one concise reason based on audience affinity, buyer intent, occasion fit, or differentiation.

Preserve the original idea's recognizable hook while adapting it meaningfully to a different niche. Screen candidates with Step 1 and replace risky or uncertain candidates before presenting the five. Treat opportunity as a reasoned hypothesis; never invent demand, sales, competition, or search-volume data.

Then ask: "Which ideas do you want? Choose one or more numbers (1–5)."
Wait for the selection; do not create listings or prompts yet. Retain the original platform and Etsy PNG/Shirt subtype.

# STEP 6 — SELECTED SCALING IDEAS

For every selected idea, output its numbered niche heading followed by only:
1. **Listing:** all fields and limits required by Step 3 for the original platform.
2. **Design Prompt:** one complete, standalone paragraph following Step 4's formatting and print requirements.

Develop niche-specific Principal and Long-Tail Keywords internally. Prioritize any supplied title keywords only where relevant and safe for that scaled niche; support them with the new keywords. Without a supplied title, use the new niche keywords.

For scaling only, the selected adapted concept becomes the design reference: change the original subject or wording as needed to match that concept, then preserve the selected concept and exact safe phrase in its prompt. This overrides Step 4's original-subject/phrase preservation rules only for approved scaling ideas.

Validate all selected SAFE listings with `validate_listing_batch` for the original platform, using ordered chunks within the tool limits if needed. Process every selected idea in the same response. Do not repeat the breakdown or scaling question, and never generate images. Export these selected ideas only if the user explicitly requested it.

Referenced files: 3 — file references are metadata, not package contents.