douyin- women-copy-production-workflow-female-custom

Douyin Copy Production Workflow - Female Edition

Use this skill for the user's complete female-facing production loop from assistant-collected scripts to publishable long-form Word drafts.

Female Edition Core Override

This skill is the complete female counterpart of the current male workflow. It has the same version set and production mechanics, but it must never share audience assumptions, prompt identity, saved sessions, or ledgers with the male edition.

Apply these rules to every direction, including 2.5, 2.6, local 2.7 fusion, 2.8, 2.9, 3.0, 3.1, and 2.5 finished-draft transplant:

• Default audience and address: women, female viewers, 姐妹们, and female students/clients.
• Default protagonist: a female student/client. Default analysis object: male psychology, male behavior, male investment, commitment, avoidance, testing, and relationship intent.
• Preserve the source topic and promised count. Convert male-workflow examples semantically: 兄弟/男性学员/男方推进/女方反馈/追女生 become 姐妹/女性学员/女方筛选或推进/男方反馈/识别和筛选男人 unless the source explicitly requires another lens.
• Ryan is the fixed female-edition persona and brand. Every identity, voice label, CTA brand, and fixed ending must use Ryan.
• Fixed ending for current Ryan long-draft directions: 我是探花Ryan，我们粉丝群里见，感谢观看.
• Keep the same sharp, dark, interest-sensitive mechanism density as the selected version. Female-facing does not mean automatically softening the copy into comfort, victim narration, or generic boundary education.
• Always prepend the stable Female Edition Prompt Override from references/gemini-expansion.md before the selected full expansion block. This is part of the female baseline, not an article-specific supplement.

Core Outcome

Produce final .docx files that contain:

1. 爆款心理学标题包装
2. 开头版本一：高阶认知课式开头
3. 开头版本二：身份点名式硬核学习开头
4. 开头版本三：保留原文开头（来自源文档）
5. 正文

Do not add risk suggestions, yellow-highlighted annotations, or [[RISKNOTE:...]] markers.

Default Paths

• Assistant source folders are often under /Users/kin/工作用（同步）/素材文稿/助理采集文案/ on Mac or D:\工作用（同步）\素材文稿\助理采集文案\ on Windows; old E:\工作用\素材文稿\助理采集文案\ is legacy-only.
• Final Word output folder for current Mac workflow: /Users/kin/工作用（同步）/7.1后双端同步文件夹
• Final Word output folder for current Windows workflow: D:\工作用（同步）\7.1后双端同步文件夹
• Legacy Windows final Word output folder: E:\工作用\素材文稿\codex工作流长文稿
• Expansion tests/comparison runs must also be exported as .docx files into the current dual-device synced workflow folder; do not deliver .txt unless the user explicitly asks for txt.
• Gemini execution choice: default to the Mac local runner unless the user explicitly asks to use Windows or the task depends on Windows-only files, old Windows Codex projects, or Windows-only tooling.
• Current Mac Gemini expansion command directory: active Codex workspace when it contains scripts/run_gemini_chat.sh; current tested workspace is /Users/kin/Documents/Codex/2026-07-02/gemini.
• Current Mac Gemini expansion command: ./scripts/run_gemini_chat.sh --prompt-file <prompt.txt> --output-file <expanded.txt>. For routine 2.5, prepend the female override, use the complete 2026-06-05 early B direct-draft block, and add --isolated for every source. If the user explicitly asks to continue a saved female 2.5 conversation, use --session-file work/gemini_session_female_25.json. Keep female 2.5-transplant, 2.8, and 2.9 in their own dedicated session files.
• TeamRouter network gate: before the first manuscript request in a task, run ./scripts/run_gemini_chat.sh --print-config --isolated and verify that TEAMO_CHAT_COMPLETIONS_URL is exactly under https://api.teamorouter.com/v1/. In restricted Codex sandboxes, run the actual manuscript request with the approved escalated network permission and the reusable prefix ./scripts/run_gemini_chat.sh; a plain sandbox request can fail DNS with Errno 8 even while TeamRouter, the API key, and the model are healthy. If this happens, rerun the identical isolated request with authorized network access. Do not change the prompt, model, or source merely because the sandbox transport was blocked. Never send the manuscript if the printed endpoint is a different host.
• The embedded Ryan 2.9 block in references/gemini-expansion.md is the female source of truth. Do not reuse the male project's external active-prompt files.
• Windows fallback Gemini expansion command directory: C:\Users\Administrator\Documents\Codex\2026-06-04\gemini3-1pro-api.
• Windows fallback Gemini expansion command: .\outputs\run_gemini_chat.cmd --prompt-file C:\path\to\prompt.txt --isolated.
• Other-copy Gemini isolation rule: for non-female-relationship copy or unrelated tests, use --prompt-file <prompt> --isolated on the selected host so saved female relationship expansion conversation state is preserved; do not use /new for this.
• Opening source of truth: invoke baokuan-kaitou-sheding / $爆款开头设定; do not hand-write openings from memory.

Reference Files

Load only the reference files needed for the current step:

• references/gemini-expansion.md: read before sending source text to Gemini, creating prompt files, retrying failures, or checking generated drafts.
• references/2.5-finished-draft-transplant.md: read when the user asks for 2.5成稿换芯, provides an already-expanded approved 2.5 draft as the style parent, or asks to preserve the mother topic while replacing its internal arguments and examples.
• references/gemini-expansion.md: also read when the user asks for 3.1, 2.5预换芯, 原稿预换芯, or wants to first rebuild the short source manuscript while keeping the exact public topic, then run routine 2.5 expansion.
• references/2.9-strong-conclusion-layer.md: approved formal 2.9 enhancement source of truth for strong conclusions, human-nature games, interest judgments, dark insight retention, and the three-topic validation record. Its prompt text is also embedded inside the formal 2.9 block in references/gemini-expansion.md so routine prompt builders cannot omit it.
• references/source-learning-and-style.md: read when processing a batch, learning source viral logic, classifying source styles, or preserving source openings.
• references/openings-and-titles.md: read after expansion passes checks and before Word export to generate openings, score them, and build title packages.
• references/export-and-hard-rules.md: read before optional scoring, Word export, final delivery, and post-run workflow updates.
• references/3.0-qianchuan-tts.md: read when the user asks for 3.0千川素材、保头压中自然转化、热门原稿植入CTA、1500字内成稿、自然过桥或直接TTS Word交付。

Workflow

1. Queue Every Provided Source By Default

When the user provides a batch of .docx scripts, do not run pre-expansion scoring, ranking, filtering, or top-N selection unless explicitly asked. Preserve stable filename order unless the user gives another order.

Read references/source-learning-and-style.md and create a concise batch learning note in the current workspace outputs folder.

2. Expand With Local Gemini

Read references/gemini-expansion.md.

Use the Mac local Gemini runner by default:

cd /Users/kin/Documents/Codex/2026-07-02/gemini
./scripts/run_gemini_chat.sh --prompt-file work/prompt.txt --isolated --output-file work/expanded.txt

Use Windows only when the user explicitly asks for Windows or the task depends on Windows-only files, old Windows projects, or Windows-only tooling:

cd C:\Users\Administrator\Documents\Codex\2026-06-04\gemini3-1pro-api
.\outputs\run_gemini_chat.cmd --prompt-file C:\path\to\prompt.txt --isolated

Build each UTF-8 prompt file with the stable Female Edition Prompt Override from references/gemini-expansion.md, then the selected full Gemini instruction block, followed by any required direction-specific injection and the original source text wrapped as 【原文开始】...【原文结束】. For routine 2.5, the official female baseline is the override + complete 2026-06-05 early B direct-draft block + source, with each source run using --isolated. Do not append the old 旧2.5风格参照 injection in routine 2.5 runs unless the user explicitly asks for a separate style-calibration test. If the user asks for 单案例、一案到底、一个案例贯穿全文, use the 2.6 Single Case direction.

If the user asks for 2.5成稿换芯, 用扩写好的2.5改内容, 保留母题替换内容, or says to keep the method approved by the August 1 fixed-mother-topic batch, use the current canonical fixed-mother-topic limited-inheritance mode in references/2.5-finished-draft-transplant.md. For every transplant mode, infer the mother topic, audience promise, central claim, and promised count from the manuscript body only. Never use the filename, folder name, Word title/core properties, document heading, stale metadata, or an earlier output title to choose the route or invent a structure; those fields are for file identification only. If any of them conflict with the body, the body wins. Prompt assembly must begin with the female override and the complete early-B 2.5 block; then append the transplant task layer, an unused route, a limited-inheritance card, cumulative route exclusions, and the approved 2.5 finished draft. Run every article through --session-file work/gemini_session_female_25_transplant_legacy.json; never use the male shortcut or male session file.

If the user asks for 3.1, 2.5预换芯, 原稿预换芯, 原稿先换芯再2.5, or asks to test the new pre-transplant mode, use 3.1 Source Pre-Transplant -> 2.5 Direct Draft from references/gemini-expansion.md. This mode is for source manuscripts before 2.5 expansion, not finished 2.5 drafts. Read and fingerprint the body before looking at identifying metadata, then keep the body-derived mother topic, audience promise, emotional direction, and any count explicitly promised by the body unchanged. A number or framing that appears only in the filename or metadata must never be imposed on the manuscript. Rebuild the source-level internal points, examples, order, and scenes; naturally embed 2-5 total topic-matched terms selected across psychology, PUA, sociology, or biology into the rebuilt manuscript itself; record the route and embedded terms in work/female_2.5_pretransplant_ledger.json; freeze that manuscript; then send it directly through the standard complete 2.5 block with --isolated. The 2-5 limit is total across all four fields, not per field, and a draft does not need to use every field. The rebuilt manuscript must start directly from its new thesis and points; never restate, summarize, mock, downgrade, or criticize the old source points as a setup for the new content. For 3.1, never append 旧2.5风格参照, an article-specific execution supplement, a safety supplement, or any other second-layer prompt. Preserve the returned 2.5 draft's original aggressive intensity by default: do not post-edit it merely to make it calmer, more professional, less provocative, less dark, or less sensational. Reject drafts that expose prompt/backstage wording such as 选题不变, 这一次我们不讲, 原稿, 旧稿, 提示词, that merely rename the old list, or that imports a count/structure found only in the filename or metadata.

If the user asks for 2.7, 融合版, 2.5和2.6融合, or wants the same effect as the approved 316 body-signal fusion draft, do not ask Gemini to write a separate 2.7 draft from the source. First generate and validate the Gemini 2.5 and 2.6 drafts, then have Codex locally fuse those two accepted drafts using the 2.7 Codex Fusion rules in references/gemini-expansion.md.

If the user asks for 2.9, 2.5×2.8融合版, 2.5和2.8融合, or 融合提示词, use the saved 2.9 Fusion Draft direction in references/gemini-expansion.md. This is a direct Gemini prompt-file direction: 2.5 supplies Ryan voice, sharp judgment, female perspective, business logic, and continuous deduction; 2.8 supplies hidden structure, concrete evidence, reduced adjective pressure, topic isolation, and stable output control. The formally approved strong-conclusion layer is part of 2.9 by default: preserve source-supported human-nature games, interest judgments, dark insights, relationship power changes, and sharp conclusions instead of automatically neutralizing them. Do not substitute the unrelated 2.7 single-case fusion flow.

Do not use the old web expansion channel. Do not send /new unless the user explicitly asks. Routine 2.5 and 3.1 are isolated per source. Female 2.5-transplant, 2.8, and 2.9 use the separate session files defined in their references. Never point a female direction at the male gemini_session.json or male session aliases.

3. Validate Each Expanded Draft

Read references/gemini-expansion.md for the exact checks and retry language.

Verify length, topic relevance, refusal/API failure text, stale-topic contamination, and required ending. For the current 2.5 baseline, the hard minimum is 4000 Chinese characters with no upper limit; 6000-8000 is a preferred depth range when the source supports it, not a cap or hard gate. Also reject 2.5 drafts whose title, hook, or central argument is built around proactive defensive qualifiers or self-negation unless the requested topic is explicitly a concept distinction. Other directions follow their own rules or the user's newest threshold.

For 2.5成稿换芯, also reject any draft that paraphrases the old list, reuses an excluded signature case, carries over old dialogue, changes the mother topic, loses the selected core source mechanism, overuses inherited material, or refers to earlier episodes/rewriting backstage. Every retained mechanism must be extended with at least two new scenes, interest shifts, consequences, or contrary proofs. Run an awkward-expression pass for invented compounds such as 位置稳的男人 or 性缩力; repair only obvious local wording defects without softening the argument. Verify the new promised count, mid-article and ending fan-group CTAs, the exact fixed ending, and the cumulative route/similarity ledger before accepting the article.

For 2.9 Fusion Draft, the user's current rule overrides that generic default: hard minimum 4000 Chinese characters, with no upper limit. Validate promised list counts, hidden structure, repeated pressure words, natural professional-term insertion, mid-article and ending fan-group CTAs, and the exact fixed ending. Also validate that source-supported strong conclusions were not softened into generic advice, every key list item contains a memorable Ryan judgment, conditions sharpen rather than cancel the conclusion, and terms such as 位置稳的男人 do not appear as awkward character labels.

4. Add Openings And Titles

Read references/openings-and-titles.md.

For every accepted article, execute these subskills in order:

1. Invoke baokuan-kaitou-sheding / $爆款开头设定 and use its approved learned opening forms as the source of truth for 开头版本一：高阶认知课式开头 and 开头版本二：身份点名式硬核学习开头. Do not hand-write generic openings from memory.
2. Extract 开头版本三：保留原文开头（来自源文档） from the source document without rewriting it.
3. Score versions one and two with the mandatory 20-point opening gate; rewrite with baokuan-kaitou-sheding until both pass.
4. Invoke viral-psych-title-wrapper / $爆款心理学标题包装器 for the title package. Generate 10 three-line candidates, strongest 3, and recommendation reason.
5. Maintain a batch title ledger to avoid duplicate mechanism chains or generic interchangeable titles.

5. Export Clean Word Documents

Read references/export-and-hard-rules.md.

Export final .docx files to the current dual-device synced workflow folder (/Users/kin/工作用（同步）/7.1后双端同步文件夹 on Mac, D:\工作用（同步）\7.1后双端同步文件夹 on Windows) with content/topic-based filenames. Do not leave generic names such as risk-note or timestamp-only filenames.

Before final delivery, verify each .docx contains the required headings and has no risk suggestions, yellow highlights, or [[RISKNOTE:...]] markers.

6. Final Reply

Keep the final answer short. Tell the user the Word file path, character count, and confirm that no risk suggestions, yellow-highlighted annotations, or [[RISKNOTE:...]] markers were added.

If optional viral potential review was requested, report it. Otherwise do not add a scoring/pass-fail statement.