x-growth-automation（x 增长自动化）

X Growth Automation

Build a separate, reusable X automation project for the user. Do not assume their current production project should be modified. Prefer creating a fresh folder unless they explicitly target an existing repo.

Core workflow

1. Ask setup questions first using references/setup-questionnaire.md.
2. Choose safe rollout mode using references/rollout-modes.md.
3. Scaffold a clean project with scripts/scaffold_x_growth_project.py.
4. Fill configs based on the user's answers.
5. Keep publish disabled by default unless the user explicitly asks for live publishing.
6. If enabling live publish, keep caps, slot scheduling, and fallback rules explicit.

What this skill should help build

A reusable X automation system with these layers:

• discovery via Bird CLI
• selection/scoring layer
• draft generation layer
• optional LLM-first approval layer
• optional source branching from an external editorial feed
• optional community CTA logic
• slot-based publishing via X API
• dry-run-first rollout

Non-negotiable safety defaults

• Create the project in a new folder unless the user clearly says otherwise.
• Treat Bird as read/discovery first.
• Treat X API as write/publish layer.
• Default new installs to dry-run.
• If the user wants aggressive automation, still keep daily/monthly caps in config.
• Never silently copy private tokens, links, handles, or niche assumptions from another project.
• Treat reply automation as a higher-risk lane than normal posts; disable by default unless explicitly requested.
• In live mode, require clear logging for posted / skipped / failed outcomes.
• Use idempotent slot keys that do not depend on mutable draft text (prevent same-source repost loops).
• Add anti-repetition guardrails (for example last-48h similarity checks + source de-dup).

Ask these questions before scaffolding

Read references/setup-questionnaire.md and ask only the missing items. Do not dump all questions at once if the user already answered some.

Minimum set to unblock setup:

• niche/topic focus
• posting language
• dry-run or live
• daily target range
• monthly hard cap
• Bird available?
• X API available?
• source branching wanted?
• community CTA wanted?

After answers are clear

Run the scaffold script:

python3 scripts/scaffold_x_growth_project.py --path <target-dir> --profile-json '<json>'

The JSON may include fields like:

• project_name
• language
• niche_summary
• daily_min
• daily_max
• monthly_cap
• community_enabled
• community_link
• reply_cta_enabled
• reply_cta_style
• source_branching_enabled
• source_branching_label
• bird_enabled
• x_api_enabled
• live_publish

Recommended setup behavior

If user says “install this system for me”

• Ask the setup questions.
• Scaffold a fresh project.
• Fill the config files.
• Keep publish disabled unless they explicitly approve live mode.
• Explain where they should place credentials.

If user says “can we adapt this to my niche?”

• Ask for niche, audience, tone, and content pillars.
• Update topic/search/watch-account config.
• Update crosspost/reply policy if needed.

If user says “make it more aggressive”

• Increase cadence and caps explicitly.
• Keep hard caps documented.
• Confirm whether replies should include CTA or not.
• Do not remove quality gates unless asked.
• Keep reply-specific safeguards even in aggressive mode: target validation, skip-on-failure, and per-day caps.

If user says “I want community integration”

• Ask whether the external community/feed is:

- source-of-truth content feed - CTA destination only - both

• Configure source branching and reply CTA rules separately.
• Never assume the community platform is Telegram unless the user explicitly says so.

File customization order

After scaffolding, customize in this order:

1. config/publish-policy.json
2. config/publish-slots.json (if slot-based cadence is used)
3. config/topics.json
4. config/budget-policy.json
5. config/draft-diversity.json (or equivalent prompt/diversity controls)
6. .env.example
7. docs/operator-notes.md
8. prompts if the user wants a distinct tone

When to enable live mode

Enable live mode only when all of these are true:

• credentials are present
• the user explicitly wants live publish
• cadence/caps are set
• reply policy is defined
• source branching behavior is defined
• the user understands what will auto-post

Publish strategy guidance

Use explicit profiles and encode them in config:

• Conservative: 2-5/day, reply lane optional, 0-2 core
• Balanced: 4-7/day, reply lane optional, 2-4 core
• High-volume: 10/day = 1 news crosspost + 1 special crosspost + 8 core posts

If the user chooses high-volume mode:

• keep monthly cap explicit
• disable reply lane unless there is a clear reason to keep it
• enforce anti-repetition guardrails (recent-post similarity + source de-dup)
• distribute core slots through the day (avoid burst clustering)

If the user wants a different mix, encode it in config rather than burying it in prose.

Resources

scripts/

• scaffold_x_growth_project.py creates a clean generic project scaffold.

references/

• setup-questionnaire.md contains the installation interview.
• rollout-modes.md explains dry-run vs live rollout choices.