Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Generate daily fortune-themed X/Twitter long posts and threads from user context. Use when the user asks for 今日运势, 运势推文, 星座, 生肖, fortune, horoscope, zodiac, daily luck, mystical copy, or reflective social content.
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 108% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 132% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 354% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 330% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 311% | 0% |
把“今日运势”从普通文案生成升级为运营级 X/Twitter 内容生产流水线。最终内容要像可发布的账号内容,而不是传统黄历摘要:有强开头、具体用户场景、反差、账号人格、互动设计、运营评分和必要的改写循环。
这个 skill 的目标不是做确定性预测,而是把“运势”包装成轻量的心理暗示、生活观察、情绪整理和行动提醒。
今日运势内容必须遵守:
本技能以西方占星 + 东方象征为底料。运行时为「日期 + 时区 + 目标星座」生成可复现的 FortuneContext(实现见 src/lib/fortune/context.ts),底料分三类,每个因子都带来源等级与置信度(sourceLevel / confidence):
目标星座:从用户输入解析(如「今日天秤座运势」→ 天秤座);未指定时回退当日太阳星座或「通用」。
Creative seeds 由日期、星座和固定映射确定性生成,只用于内容多样性和创作主轴,不是命理事实,不得写成「星象证明今天一定会怎样」。今天写什么主轴由 FortuneContext.creative.focusDomain 提供方向,但它是创意种子(creative-rotation),不是预测结论。
解读规则见 references:astrology-signs.md(十二星座底料)、astrology-daily-engine.md(月相/星期/侧重域 → 当日基调)、eastern-symbolic-calendar.md(五行/节气/生肖 → 现代行动)。
象征只是情绪与行为的隐喻,不是命运开关(遵守 Safety Positioning)。
象征系统(星座、月相、五行、节气、塔罗)只是镜子。本技能以 Seth 式意识框架作为意义解释内核(写作透镜,不是预测引擎),必须读取 references/seth-consciousness-framework.md,把象征翻译成:
要求:
Safety Positioning 和 Seth Consciousness Layer 都是内部审查与解释逻辑,不是 reader-facing 正文内容。最终 final.longTweet.body / final.thread 必须像自然发布的 X/Twitter 运势内容,不能像 AI 安全报告或心理学说明书(解读规则见 references/public-post-boundary.md、references/playful-fortune-voice.md)。
final.longTweet.body / final.thread 禁止出现:把这条当作、仅供娱乐、娱乐与反思、不是预测、这不是预言、不构成投资建议、不保证、安全提醒、风险提示、审查、Seth、pipeline、情绪不是命令、概率线不是固定。reviewNotes,不写进正文。Use this skill when the user asks for:
运行时把这套流程编排为 5 段独立的模型推理(understand → diverge → judge → draft → refine),而不是一次成稿;角度判分与改写由独立的「运营主编」段真实执行(实现见 src/lib/fortune/pipeline.ts)。下面的步骤描述每段应产出的内容与质量标准。
Do not immediately write the final post. Produce the artifact through this sequence:
draftV1.draftV1 as an X/Twitter operator editor.final.finalize_twitter_creative with the full dailyFortune artifact.Before writing, identify the target user's hidden pain and emotional need.
For “海外中文年轻人”, include at least two concrete scenes unless the user explicitly asks for generic content. Consider:
Final content must include at least 2 real scenes from the audience's life. Do not leave these only in analysis.
Generate 3-5 angle options before final writing.
Each option must include:
anglethesisemotionalHookconcreteScenewhyItWorkssafetyRiskGood angle patterns:
Choose one angle with the strongest mix of specificity, emotional pull, and safety.
Generate 5 hook options before final writing:
Each hook option must be an object with:
type: contrarian | scene | confession | mystical-image | practical-warningtextwhyItWorksAvoid generic openings:
These are only acceptable if followed by a strong contrast in the same opening sentence.
Build a concise spine:
When the user requests longTweet, the final body must be 600-1200 Chinese characters.
It must include:
For longTweet, final.longTweet.body is the primary artifact. The top-level tweet field is only a short summary and must not replace the long post.
When the user requests thread, output 5-8 tweets.
Each thread item must have one of these roles:
For thread, final.thread is the primary artifact. Each item should be independently readable and suitable for copying into X/Twitter as a thread.
After draftV1, score it as an X/Twitter operator editor:
Each score is 1-5. If any score is below 4, rewrite once. The final artifact should have all scores at least 4 unless safety requires a publishability downgrade; explain any exception in operatorCritique.problems.
Final content must include one natural interaction design:
Avoid spammy CTA such as “转发立刻发财” or “关注马上转运”.
Preferred phrases:
Avoid:
Return valid JSON in dailyFortune, then call finalize_twitter_creative.
json{ "selectedSkill": "daily-fortune-tweet", "outputType": "longTweet | thread | both", "inputSummary": { "topic": "string", "audience": "string", "tone": "string", "assumptions": ["string"] }, "audienceInsight": { "corePain": "string", "realScenes": ["string"], "emotionalNeed": "string" }, "angleOptions": [ { "angle": "string", "thesis": "string", "emotionalHook": "string", "concreteScene": "string", "whyItWorks": "string", "safetyRisk": "string" } ], "selectedAngle": { "angle": "string", "reason": "string" }, "hookOptions": [ { "type": "contrarian | scene | confession | mystical-image | practical-warning", "text": "string", "whyItWorks": "string" } ], "fortuneSpine": { "keyword": "string", "symbolicImage": "string", "audienceSpecificScene": "string", "emotionalWeather": "string", "coreTension": "string", "practicalAdvice": "string", "tinyRitual": "string", "closingImage": "string" }, "draftV1": { "longTweet": "string", "thread": [ { "index": 1, "text": "string", "role": "string" } ] }, "operatorCritique": { "hookStrength": 1, "specificity": 1, "audienceFit": 1, "emotionalResonance": 1, "shareability": 1, "saveWorthiness": 1, "safety": 1, "problems": ["string"], "rewriteDirection": "string" }, "final": { "longTweet": { "title": "string", "body": "string", "hashtags": ["string"] }, "thread": [ { "index": 1, "text": "string", "role": "string" } ] }, "engagementPlan": { "cta": "string", "commentPrompt": "string", "seriesLabel": "string" }, "reviewNotes": { "safetyCheck": ["string"], "hypeCheck": ["string"], "publishReadiness": "draft | reviewed | publish-ready" } }
If the user only asks for longTweet, return final.thread: []. If the user only asks for thread, return final.longTweet.body: "". If both are requested, return both.
operatorCritique present with all scores 4 or above after rewrite?engagementPlan present and non-spammy?final.longTweet.body the primary full content?final.thread have 5-8 items with clear roles?Other measured skills in the registry, with their headline benchmark lift.