ezoic-draft-content

Ezoic — Draft Content

Write the final copy for a single channel asset, in Ezoic's voice, grounded in canonical sources.

Step 0 — Boot

1. Read ${CLAUDE_PLUGIN_ROOT}/references/_resolver.md first. The 3-tier resolver and the 12 + 6 brand-enforcement contract are binding as written there.
2. Load from ${CLAUDE_PLUGIN_ROOT}/references/: brand-voice.md, channel-playbook.md, product-catalog.md, proof-points.md, ezoic-context.md. If the asset will be seen (social image, ad), also load ezoic-style-design-bible.md and check the live asset catalog.

Step 1 — Confirm the brief

Use the brief block from the orchestrator. If invoked directly, build the 10-field brief (Deliverable, Channel, Audience/ICP, Author, Pillar, Narrative frame, Primary proof, Value-prop rung, Voice calibration, Length target). If a field can't be filled from canonical sources, stop and ask. Never guess ICP, pillar, or proof point.

Step 2 — Produce the copy

Pull the channel's structure, length target, and voice calibration from channel-playbook.md (Press Release, LinkedIn Post, Blog/Website Copy, Help Center FAQ, etc.). Match the output to exactly what was asked — copy only unless an image or extra format was requested. Don't format-vomit.

Per-channel anchors:

• Press release: headline = news + directional signal; dateline; lead ¶ then "why" ¶; one attributed exec quote that carries the narrative (not the headline); proof paragraphs; boilerplate. 400–600 words.
• LinkedIn (exec): first line stands alone; short paragraphs, one idea each; pivot to Ezoic's focus; one or two specifics; no "Sign up today" CTA. 150–300 words.
• Blog / product page: lead with the tension or the value prop; one claim per section; proof from proof-points.md.
• FAQ: lead with the answer, then minimal rationale; no marketing language.

Step 3 — Visual direction (only if the asset will be seen)

Apply visual rules 13–18 from _resolver.md. Place real assets only from the live catalog; never infer an asset from its filename.

Step 4 — Self-check

Run the 10-point smell test from _resolver.md verbatim. Then enforce the skill-specific guardrails:

• One narrative frame per asset — don't stack the five strategic narratives.
• Banned words auto-rewrite — strip and replace per brand-voice.md → Anti-patterns.
• No CPM dollar claims without a cited source.
• At least one proof point with its required segment context appears in the first third.

Fix anything that fails, then deliver.