TranFu Product Marketing Outline Skill

中文

build-marketing-outline is a TranFu-specific Codex Skill. It inspects a product URL, an unpublished product repository, or both, then turns product evidence into a formal launch-marketing strategy and editor-level outlines for six channels.

It produces strategy and outlines only. It does not create finished copy, images, platform drafts, or publishing actions.

This version includes the tranfu.com and Agent 公司养成记 narrative, uses Chinese output, the Asia/Shanghai time zone, and six Chinese content channels, and defaults to ~/Documents/product-marketing as its output root. It is not a general-purpose marketing Skill.

Core Capabilities

• Supports a public product URL, an absolute project directory, the current project directory, or a URL and source repository together.
• Separates page evidence, existing results, end-to-end tests, source implementation, inference, hypothesis, and conflict.
• Keeps formal launch marketing as the target for unpublished products while naming every release blocker.
• Produces channel-specific outlines for Xiaohongshu, WeChat Official Accounts, Zhihu, Toutiao, Weibo, and the product website.
• Produces a nine-section, platform-neutral master outline for a downstream content-production workflow.
• Keeps the product read-only and does not install dependencies, change source files, or trigger product-side writes.

Installation

Install from the company Skill catalog (recommended):

tfs install build-marketing-outline --scope user

Install from the standalone GitHub repository:

mkdir -p ~/.codex/skills
git clone https://github.com/BruceL017/build-marketing-outline.git \
  ~/.codex/skills/build-marketing-outline

Update:

git -C ~/.codex/skills/build-marketing-outline pull

After installation, start a new Codex task or let Codex refresh its available Skills.

Runtime Requirements

• A Codex environment with Skills support.
• URL mode requires network access. With browser capabilities it can inspect runtime behavior; otherwise it falls back to HTML evidence and lowers the release verdict.
• Project mode requires source-read access. Git is used to compare repository state, and local runtime inspection uses only already-installed dependencies and documented commands.
• Enable local runtime inspection only for trusted projects. It may execute the project's own development command and is not an independent code sandbox.

Usage

Analyze a product URL:

$build-marketing-outline https://example.com

Analyze an unpublished product project:

$build-marketing-outline /absolute/project/path

Run from the product-project root:

$build-marketing-outline

Combine runtime and source evidence:

$build-marketing-outline https://example.com /absolute/project/path

You may also specify an output root in the request. The output location must remain outside the inspected product project.

Provide only credential-free public URLs. A URL containing userinfo, a token, API key, signature, authorization value, or login code is rejected and is not fetched or written to the output.

Output

Each run creates a new timestamped directory:

{product-name}_营销内容大纲执行包_{YYYYMMDD-HHmm}/

A normal run contains exactly six documents:

1. 00_执行摘要与发布判定.md
2. 01_产品证据与用户路径.md
3. 02_营销策略与内容矩阵.md
4. 03_六渠道编辑大纲.md
5. 04_通用母大纲.md
6. 05_素材清单与发布门禁.md

The Skill produces only a diagnostic document when the input is completely inaccessible or contains no identifiable product evidence.

Evidence Levels

An HTTP 200 response, page load, rendered shell, or navigation-only check is E0-P, not E0-T. E0-S proves that a project implements something; by itself, it does not prove that users can use it now.

Read-Only And Safety Boundaries

The Skill does not automatically:

• Install or upgrade dependencies.
• Modify product files, configuration, or databases.
• Submit forms, upload files, save settings, or clear data.
• Invoke AI generation, grading, payment, email, messaging, or publishing APIs.
• Log into user or administrator accounts.
• Read .env*, secrets, real user data, database content, or production logs.
• Perform external competitor, keyword, trend, pricing, or market-size research.

In project mode, it compares the complete Git status before and after inspection and stops any local server it started.

Generated documents preserve credential-free input URLs and local absolute evidence paths. Before sharing an execution pack externally, review and redact local usernames, directory structure, and other internal paths.

Release Verdicts

The Skill uses four verdicts:

• 可进入发布制作 - ready to enter copy and asset production
• 待补证据或产品条件 - waiting for evidence or product conditions
• 仅内部使用 - internal use only
• 无法形成可信大纲 - unable to form a credible outline

The first three verdicts retain a formal-launch target. The Skill does not silently replace the campaign with prelaunch content or a build log.

Repository Structure

.
├── SKILL.md
├── agents/
│   └── openai.yaml
├── references/
│   ├── evidence-and-readiness.md
│   ├── output-contract.md
│   └── tranfu-marketing-framework.md
├── README.md
└── README.zh.md

Validation Status

• The core Skill content passed Codex quick_validate.py in the standalone source repository. This directory uses company catalog frontmatter and is checked by the tranfu-skills validators.
• Passed isolated forward tests for both URL and project-directory modes.
• Verified the six-document contract, six channel outlines, evidence IDs, TranFu return path, and release gates.
• Verified that project-mode testing leaves the product repository and local services unchanged.

License

The governing license is the LICENSE file at the root of the repository containing this Skill. If that repository has no LICENSE, public visibility alone does not grant permission to copy, modify, or distribute it.