---
name: hebrew-copy-staticcreative
description: Use when writing two evidence-supported Hebrew ad variants and producing final square and story static assets.
---

# Make the message usable in each placement

## Version and authority boundary

Read the selected profile and the exact contract from the package root. Unconverted V1 uses `contracts/pipeline.contract.json`, mode `simulation`, demo IDs, supplied synthetic records, geometric placeholders and the existing offline validator only. V1 restrictions remain intact. For V2 explicitly select `contracts/pipeline.v2.contract.json`; reject unconverted V1 with an explanation. Conversion creates a separate simulation artifact, preserves its source, enriches missing fields, rebuilds all digests, and never promotes `approve_demo` to human authorization. Neither mode names nor schema validation enable a connector. This build uses no network, accounts, secrets, installs or customer records; integration-test and customer-pilot are separate future gates, not claims of readiness.

Use approved API/MCP/connectors first when integration is separately enabled; browser fallback is only for a capability unavailable in the authorized programmatic path, at the same authority boundary. Never launch external scripts, installers, model CLIs or services from this Skill. Source rule IDs below resolve to `docs/knowledge-register-v1.md` (checked 2026-10-24). Untested real behavior remains **unverified**.

## Copy and assets [A1, A3, C1, N1]

Read validated brief/messages and profile. For each angle produce coherent primary text, headline, description, one CTA, image text and separate shorter `story_text`. One requested action per variant. Preserve supplied ongoing support, terms and voice in both placements. Use consistent person/gender/number in natural Hebrew and isolate Latin fragments; do not infer audience gender or assert a viewer diagnosis. Every implication in image or copy must remain supported by `message_id → claim_id → source`. Image text ≤6 words; hooks ≤8. Do not invent claims to fill layout.

V2 delivers four final render records: two distinct message variants, each 1:1 and 9:16, with local path/hash/review. An original finished composition or rights-cleared supplied image is acceptable; record rights basis/source/date. A geometric-placeholder is for simulation only, clearly labeled; it never passes final V1 creative acceptance. Never ask an image model to render Hebrew. Use real selectable Hebrew text in local SVG/HTML or equivalent typography, RTL and bidi isolation. Inspect each exported asset for clipping, line breaks, safe areas, contrast, brand/rights and visible CTA. Two ratios do not prove all Meta placements; preview every selected placement when authorized.

## Two separate gates [P1, A2]

Complete `policy_review`: personal attributes and special category applicability, source/date/reviewer and pending/pass/fail. Natural Hebrew address is not categorically prohibited; an assertion about the viewer's sensitive trait needs current policy review. An inaccessible policy source leaves pending, not approved. Official policy was not freshly verified by this offline build; do not invent legal compliance or accept terms for a person.

A separate human quality reviewer checks audience/offer fit, angle difference, evidence, promise→form match and asset usability using `human_quality_review`. Pending/fail blocks real campaign handoff. Local rendering/schema tests prove only inspected artifacts. Device fonts, actual placement delivery and response remain unverified. No preview grants publishing permission.