copywriting（文案写作）

Copywriting

Two modes: Write (new copy from scratch) and Edit (audit and rewrite existing copy). Auto-detect which applies — if copy already exists, default to Edit. Confirm with the user if ambiguous.

Reference files

Mode A — Writing new copy

Writing progress:
- [ ] Step 1: Gather context
- [ ] Step 2: Lock the brief (hard gate)
- [ ] Step 3: Discover brand voice
- [ ] Step 4: Choose framework and load references
- [ ] Step 5: Write 2–3 alternatives
- [ ] Step 6: Recommend and explain

Step 1 — Gather context

Ask these four questions before writing a single word. Do not proceed until all four are answered.

1. Page purpose — What is the one action this page must drive? (e.g. sign up, book a demo, download)
2. Audience — Who is the specific reader? Job title, pain, what they've already tried.
3. Product — What does this product do, and what is the concrete outcome for the user?
4. Traffic source — Where is the reader coming from? (cold ad, warm email, organic search, referral)

Traffic source determines temperature. Cold traffic needs more Why. Warm traffic can lead with How or What.

Step 2 — Lock the brief (hard gate)

Before writing, state the brief back to the user and get explicit confirmation:

Brief:
- Page: [page type]
- Goal: [single action]
- Reader: [specific audience]
- Core outcome: [what changes for the reader]
- Tone: [inferred from brand voice or user-stated]
- Traffic temperature: [cold / warm / hot]

Confirm this is correct before I write.

Do not write copy until the user confirms. If they push back on any point, update the brief and re-confirm.

Step 3 — Discover brand voice

Look for brand voice signals before inventing one:

• Read existing copy files, README headers, or marketing pages if accessible
• Ask: "Do you have brand guidelines, a tone-of-voice doc, or existing copy I should match?"
• If nothing exists, infer from the product and audience: B2B SaaS defaults to direct and confident; consumer apps can be warmer; developer tools lean terse and honest

Note the inferred voice in the brief. Never default to generic corporate warmth.

Step 4 — Choose framework and load references

Load references/frameworks.md and references/page-types.md.

Choose the primary framework for this copy based on the brief:

You can layer frameworks. Why/How/What almost always applies to hero copy regardless of the primary choice.

Step 5 — Write 2–3 alternatives

Write exactly 2–3 distinct alternatives. Each must:

• Apply the chosen framework visibly
• Lead with Why, not What
• Use no banned words (see below)
• Include a headline, subhead, and at least one CTA per alternative
• Be structurally different — not the same idea with different adjectives

Label each: Option A, Option B, Option C.

Step 6 — Recommend and explain

Pick one option and state clearly which and why in one sentence. Give the user a specific edit note for each alternative they did not pick — what it would take to make it stronger.

Mode B — Editing existing copy

Editing progress:
- [ ] Step 1: Read all copy-bearing files
- [ ] Step 2: Set the north star
- [ ] Step 3: Audit against persuasion frameworks
- [ ] Step 4: Remove AI writing patterns
- [ ] Step 5: Run seven sweeps
- [ ] Step 6: Flag weakest elements with labels
- [ ] Step 7: Rewrite flagged sections
- [ ] Step 8: Output before/after diff

Step 1 — Read all copy-bearing files

Scan for all reader-facing text: README headers, landing page components, hero text, CTAs, product descriptions, feature lists, onboarding strings, meta descriptions, email subjects.

Ask which files to target if unclear. Never audit copy you haven't read in full context.

Step 2 — Set the north star

Write one sentence before auditing anything: "[User] can now [do X] without [old pain]."

Every flagged line and rewrite must serve this sentence. If you cannot write it confidently, ask the user — the copy will be unfixable until the value proposition is clear.

Step 3 — Audit against persuasion frameworks

Load references/frameworks.md. Check every major copy block against each framework. Mark candidates for flagging. Do not flag everything — identify the 3–7 highest-impact problems only.

Step 4 — Remove AI writing patterns

Load references/word-lists.md and references/ai-patterns.md.

Scan for AI-isms and flag each with [AI-ISM] and the specific pattern type:

• Tier 1 words — always flag and replace (delve, leverage, robust, seamless, paradigm, holistic, etc.)
• Tier 2 clusters — flag when 2+ appear in the same paragraph (harness, empower, streamline, elevate, etc.)
• Structural patterns — formulaic openings, chatbot artefacts, "let's" transitions, significance inflation, copula avoidance

Skip this step if the user asked specifically for persuasion-only editing. Run this step first (before sweeps) if the user asked specifically for AI pattern removal.

Step 5 — Run seven sweeps

Load references/sweeps.md and run all seven sweeps in order. Do not skip sweeps because the copy "looks fine" — each sweep targets a distinct failure mode.

Step 6 — Flag weakest elements

Attach a label inline to every weak line. Use exactly these labels:

Flag the 3–7 weakest elements. Prioritise by impact on conversion or comprehension.

Step 7 — Rewrite flagged sections

Rewrite each flagged block:

• Lead with Why (the user's problem or desire), not What (the product)
• Name the concrete outcome, not the capability
• Replace adjectives with proof: instead of "powerful analytics", write "see which pages kill signups"
• Make CTAs outcome-specific: "Start syncing" beats "Get started"
• Every sentence must add new information or it gets cut
• Never lengthen a CTA in the name of clarity
• When replacing AI-isms, rewrite the sentence — don't just swap the flagged word for a synonym

Step 8 — Output before/after diff

## Copy Audit — [file or component name]

**North star:** [one-sentence value prop]

---

### [Section name]

**Before:**
> [original text]

**Issues:** `[LABEL]`, `[LABEL]`

**After:**
> [rewritten text]

**Why:** [one sentence explaining the change]

---

### Summary
- N issues flagged across N sections
- Top pattern: [most common label]
- Confidence: [high / medium — note if copy context was limited]

Banned words

Never write these — flag immediately in edit mode. Full replacement list in references/word-lists.md.

delve, leverage (verb), robust, seamless, holistic, paradigm, game-changing, cutting-edge, innovative, synergy, revolutionary, effortless, world-class, powerful

Also ban "simple" used as a claim ("our simple onboarding") — never earned upfront, always reads as a promise not yet kept.

Gotchas

• Do not write before the brief is confirmed. Skipping Step 2 produces copy that sounds good but serves the wrong goal.
• Do not flag every line in edit mode. Over-flagging dilutes attention and makes the audit useless. 3–7 issues, prioritised by impact.
• Do not audit copy you haven't read in context. A vague-looking line may be intentional contrast with adjacent copy.
• Do not impose locale changes. Don't switch British to American spelling (or vice versa) unless the project already uses the target locale — check existing copy first.
• Do not lengthen CTAs. "Sync your data across every device instantly" is worse than "Start syncing".
• If the product name appears in the first hero sentence, it's almost always [WHAT-NOT-WHY].
• "Simple", "easy", "powerful", and "seamless" are automatic [TELL-NOT-SHOW] — they are never earned upfront.
• Do not generate copy with an unconfirmed value proposition. If you cannot write the north star sentence, stop and ask.
• CTAs often live in components, not markdown. Ask the user where CTAs live if not obvious from the files.
• Traffic temperature changes the entire strategy. Cold traffic from a paid ad needs far more Why than warm traffic from a referral link. Ask if unsure.
• Brand voice mismatch is the fastest way to make good copy feel wrong. Always check for existing copy before inventing a tone.

Skill handoffs