Title Generation for WeChat Articles: Hook Levels and Candidates

Title suggestion is a handoff, not local generation: the CLI prepares a scored prompt, and the host agent's model runs it. Full reference.

The command

md2wechat title suggest article.md --json

Flags: --target-reader, --count (8-10, default 10), --max-title-chars (default 25), --hook-level 1|2|3, --prompt (default wechat-title-expert).

The handoff contract

The response is TITLE_SUGGEST_REQUEST_READY with status: action_required and execution_owner: host_agent. It is recommendation_only — no side effects, nothing written. The bundled wechat-title-expert prompt scores candidates as: 好标题 = 好奇缺口 × 价值预期 × 情绪驱动力 − 失望风险, and must return 8-10 candidates across at least 4 strategies with at least 2 direct-value titles.

Three hook levels

• Level 1 restrained: factual, low tension. Good for professional columns.
• Level 2 punchy: stronger pull, still honest. Good for tutorials and roundups.
• Level 3 high_tension: maximum curiosity gap. Use sparingly — disappointment risk grows with tension.

Landing the final title

WeChat enforces 32 characters. Apply the pick at conversion time:

md2wechat convert article.md --preview --title "Your final title"

or put it in frontmatter. Pair with Cover Image to finish the draft package.