- name
SKILL.md
The skill file stays in the publisher's repository.
Paste the copied command in a terminal. It installs this SKILL.md into that agent.
npx skills add backtrue/thinkwithblack-skill --skill pmax-copy -a cursor -yPaste the copied command in a terminal. It installs this SKILL.md into that agent.
npx skills add backtrue/thinkwithblack-skill --skill pmax-copy -a claude-code -yPaste the copied command in a terminal. It installs this SKILL.md into that agent.
npx skills add backtrue/thinkwithblack-skill --skill pmax-copy -a codex -yPaste the copied command in a terminal. It installs this SKILL.md into that agent.
npx skills add backtrue/thinkwithblack-skill --skill pmax-copy -a grok -ySend the copied message in a Grok Bot chat. The Bot fetches this SKILL.md from GitHub and adds it to the skills all your Bots share.
Install this skill: https://github.com/backtrue/thinkwithblack-skill/tree/HEAD/en/extensions/pmax-copyFrom the publisher
Work in this order: “entry point → need scenario → landing-page support → role-based expressions of the same scenario → asset pool.” Use S / P to prompt recognition and verifiable E to support reasons for choosing. Give each line one task; keep the entire copy set aligned with one scenario and brand solution
Start from Persona × USP and basic E. Retain the approved strategy and complete this skill's entry-point, scenario, and page mapping; a pure rewrite does not rerun the account inventory or require reapproval. Existing performance data is not a prerequisite for new copy
First read Copy allocation and delivery checks. That document maintains asset tasks, bridge options, the character contract, and review questions; this page sets the execution order. Review existing copy by working back through step 1 below as well; adding tone labels or recounting characters alone is insufficient
Record any lack of firsthand customer service / consultation data honestly; verified pages can still support creative hypotheses, without inventing customer quotations, demographics, or preferences. If evidence is missing for one claim, remove or narrow it and continue supported copy; ask only when the audience, product, or core facts cannot be established
See Public Skill Rules for general working principles. This skill can be used independently; its method draws on matching claims with evidence, translating features into benefits, FABE, and benefit-led headlines
For writing, rewriting, review, and replacement copy needed to meet minimum asset counts, first run the Health personalization policy check. Read the user's policy flags, avoid-use list, or screenshots for the same brand, then check the current official rules for the service, landing page, and targeting. Establish the writable scope for this round before translating scenarios into role-based expressions
Do not treat flagged text or media as usable again merely by changing synonyms, resizing, reuploading, or changing the wording to “explore / learn about / physician assessment.” Persona and pain points are internal hypotheses only; they cannot justify claims about the reader's health, intimate concerns, or bodily flaws. Keep three sources distinct in policy interpretation: “account flags,” “official rules,” and “editorial judgment for this round”
If account records are unavailable, use the supplied products, pages, and official policies to complete drafts and item-by-item checks, listing claims still requiring verification. When the issue concerns the landing page or targeting, list the specific items to address; do not treat a copy replacement as removal of restrictions or change the website or audience yourself
Write one brief for each set of 25 copy assets:
One person can have different entry points; one entry point can have different scenarios. If a set focuses on different body areas, objections, or usage tasks at once, split it into scenario drafts first, or choose the main scenario using the approved strategy, then write each asset pool. This is message planning and does not automatically add or reorganize account campaigns / asset groups
Read the page before making ad promises. If the first screen does not match the entry point, record the gap and complete copy using content the page actually supports. Do not invent exclusive offers, comparison tables, or services, or change the website yourself. E explains why the brand can meet this scenario; listing numbers, applicators, or addresses alone does not complete persuasion
Keep the entry point, scenario, core solution, and evidence boundaries from step 1 fixed, and write one communication direction each for S / N / T / F:
Check that all four sentences still answer the same question before expanding the asset pool. S about the eye area, N about the whole face, T listing equipment specifications alone, and F switching to fear of pain do not count as completed role translation merely because all labels are present. Roles may choose different supporting facts for the same solution but must not replace the main need scenario
Keep “course tones S / N / T / F” separate from “SPIN S / P”; the former are ways of expressing, while the latter are situation and problem. Neither determines the reader's personality or health status
Write each asset according to the allocation in the reference document:
Use relevant E to support the main persuasive content of long headlines and descriptions. Translate features into benefits or decision reasons useful in this scenario, avoiding a whole set built around “Do you want → check the specifications first → book a consultation.” If evidence supports only selection criteria, write selection criteria rather than promoting them into treatment guarantees
Each line must make sense alone and combine with lines in other roles; do not assume H01 is followed by L01 and then D01. Independence does not require every line to contain the brand, scenario, specifications, evidence, and CTA. Use the entire pool to fulfill the tasks; avoid repeating the brand or adding vague modifiers to fill character counts
First check each asset against policy records and the avoid-use list, withdrawing matches and recording the source; then use the reference document to review the set's entry point, scenario, page, and role translation, followed by each asset's task, benefit, evidence, click reason, and combinations. Return to the brief for mixed scenarios; rewrite role expressions when labels differ but content does not; return to feature-to-use mapping for specifications alone; add genuinely new information when descriptions repeat headlines
Record “which line / group, the original problem, and the actual change.” When reviewing existing copy, retain necessary before-and-after comparisons; do not merely mark everything passed or assign yourself high scores. Without performance data, explain the reason it may attract interest without claiming CTR will increase
For claim review, separately record source location, applicable scope, inference, and unknowns. Do not elevate clinic / brand statements into independent outcome verification; a fact about one body area cannot extend to an entire product / treatment. Do not add fear, body shaming, fake testimonials, or unsupported promises for clicks
After semantic review, proofread the existing outward-facing copy in Taiwan Traditional Chinese and recount characters. Fluency does not replace a persuasion review
Deliver tables in the conversation and verify quantity, task allocation, tone allocation, and weighted characters against the reference document. Count full-width Chinese characters as 2 and half-width English letters / numbers as 1, including spaces and punctuation; verify special glyphs separately. List the count for each asset and recheck after revisions
Deliver the entry / scenario / landing-page brief, four role-based communication directions, all 25 assets with weighted counts, tasks, tones, SPIN, and E mappings; label the fifth lines' bridge functions and include the actual revision log. Include character counts, claim support, and before-and-after comparisons
Before uploading, read current Google text-asset guidance. If the user's writing range conflicts with platform requirements, follow the reference document to separate the review version from concise platform replacement copy, explicitly showing replacement IDs and differences; do not silently change the user's character requirements or label the review version platform-ready
After copy review, deliver the approved text and asset requirements. Media matching, missing-asset briefs / storyboards, and account operations are handled separately by advertising execution work; generated illustrative visuals are not evidence of equipment or outcomes
Produce, rewrite, and review Google PMAX text assets. Use when the user requests PMAX ads, headlines and descriptions, stronger reasons to click, or a complete copy rewrite. Start from Persona, USP, a
---name: pmax-copydescription: Produce, rewrite, and review Google PMAX text assets. Use when the user requests PMAX ads, headlines and…---# en/extensions/pmax-copy/SKILL.mdOverview
Produce, rewrite, and review Google PMAX text assets. Use when the user requests PMAX ads, headlines and descriptions, stronger reasons to click, or a complete copy rewrite. Start from Persona, USP, and evidence; establish the entry point, need scenario, and landing page, then express the same scenario through S / N / T / F. Deliver a 15 / 5 / 5 pool of independently recombinable assets with item-by-item review. Match images and videos only after copy completion; this skill does not cover RSA formatting or ad account operations
FAQ
Download en/extensions/pmax-copy/SKILL.md from the publisher, or copy the path and open the repository. PlugMCP lists the skill and does not host the file.
The publisher keeps it at https://github.com/backtrue/thinkwithblack-skill.
Claude, Cursor, Codex, and Grok Build can read a SKILL.md you add from the publisher's repository. Grok Bot installs it when you send the skill's GitHub folder link in a chat.