Ai hotdog

AI-HOTDog

AI-HOTDog is a configurable hotspot radar for turning media, search, social discussion, and vertical communities into cited content ideas. AI is the default profile, but the workflow must support any user-selected domain such as finance, education, food service, real estate, gaming, parenting, or a custom topic.

Operating Principles

• Treat the user's chosen topic as configuration, not as hardcoded AI coverage.
• Initialization is an interview, not a silent script run. When required setup fields are missing, ask the user first.
• Prefer stable public sources for the main evidence chain. Use Chrome or Computer Use login-state platforms only as supplemental discussion signals unless a source is an official/original page.
• Never save passwords, OTPs, cookies, local storage, session stores, or browser profile data.
• When login, CAPTCHA, risk control, payment wall, or permission prompts block access, hand off to the user and mark the platform status. Do not bypass restrictions.
• Never silently initialize a new workspace. If .ai-hotdog/config.json is missing or incomplete, ask the setup questions first, then create the config.
• Treat account authorization as a required pre-scan gate for selected login-state sources. The user may bind now, skip selected platforms, or continue with public sources only.
• Do not put any item into 热点 Top 10 unless it has a primary source URL, source type, capture time, and confidence label.
• Do not use social discussion as the only evidence for a major factual claim. Pair it with an official, original, media, code, paper, or search/news source.
• Every report must show scanned source counts, usable source counts, candidate counts, scored counts, final cited counts, and source-share percentages.

Command Router

初始化 AI-HOTDog

Create or update the workspace radar configuration. Default state directory: .ai-hotdog/ in the current workspace unless the user provides another path.

If no valid .ai-hotdog/config.json exists, the first response must be a setup question. Do not run with empty defaults. Do not create a config until the user has answered or explicitly accepted defaults. Start with this compact setup dialogue:

先初始化 AI-HOTDog。我需要确认 5 件事：
1. 你要关注什么主题/行业？例如 AI、财经、教育、餐饮、游戏，默认 AI。
2. 你希望它自动做什么动作？例如每日热点报告、自媒体选题、竞品监控、行业简报。
3. 你希望多久运行一次？手动、每天、每周，或者指定时间。
4. 你要覆盖哪些平台？公开搜索/媒体、X、YouTube、知乎、微博、B站、小红书、Reddit、LinkedIn。
5. 登录态平台要现在授权吗？可以选择现在绑定、稍后绑定，或本次只跑公开源。

If the user says "用默认 AI 配置" or accepts defaults, set the default AI profile and still ask whether login-state platforms should be authorized now, skipped for this run, or left for later.

Ask for missing high-impact configuration in a compact way:

• focus_topic: topic/domain to monitor. Default to the AI profile only when the user asks for AI or accepts the default.
• core_keywords: required search terms.
• exclude_keywords: terms to filter out.
• entity_keywords: brands, companies, people, products, or communities to track.
• regions: global, China, or specific countries/regions.
• content_goal: content ideas, industry intelligence, product opportunities, competitor monitoring, or custom.
• automation_action: the action the automation should perform, such as daily hotspot report, content ideas, competitor monitoring, or industry brief.
• automation_schedule: manual, daily, weekly, or a user-provided time description. Store the human-readable schedule, but leave actual scheduling to Codex automation.
• platform_scope: stable public sources, login-state platforms, or both.
• login_sources: selected login-state platforms that should be checked before scans.
• auth_policy: prompt each interactive run, skip unavailable login platforms, or public sources only.
• frequency: manual, daily, or weekly.
• output_language: default to the user's language.

Use references/source-registry.md to build the default source set and source mix buckets. Save only non-secret configuration, source choices, and status metadata.

For deterministic setup, run:

python3 scripts/init_ai_hotdog.py --workspace <workspace> --topic "<topic>" --profile ai --core-keywords "<keywords>" --automation-action "daily_hotspot_report"

Read references/config-schema.md before creating or editing .ai-hotdog/config.json or .ai-hotdog/source-status.json.

After initialization, immediately summarize:

• selected topic and action
• selected schedule/frequency
• selected public and login-state platforms
• whether account authorization is required
• the next recommended command

If any selected source has access_mode=chrome_login or access_mode=manual, offer to run 授权登录 AI-HOTDog before the first report. If the user skips, keep those sources in login_required or manual_required and continue with public sources only.

授权登录 AI-HOTDog

Use this immediately after initialization or before a report run when login-state sources are selected.

Read references/login-and-access.md. For each selected login-state platform:

1. Open the platform in Chrome.
2. Check visible login state only.
3. If logged in, run a safe search with the configured keywords and record available or limited.
4. If not logged in, ask the user to complete login or QR-code scan in Chrome.
5. After the user says it is complete, recheck visible access.
6. If the user skips or verification remains blocked, record login_required, manual_required, or limited and continue with the next platform.

绑定 AI-HOTDog 账号

Use this when the user wants to connect logged-in platforms such as X, YouTube, Zhihu, Weibo, Bilibili, Xiaohongshu, Reddit, or LinkedIn.

Read references/login-and-access.md before using browser tools. Open each requested platform in Chrome, let the user perform login or verification, then check whether the platform can search, open result details, and expose usable public URLs or visible titles. Record the platform status in .ai-hotdog/source-status.json or the user's configured status file.

重新绑定 AI-HOTDog 账号

Use this when a platform is login_required, manual_required, limited, or unavailable. If the user names a platform, only rebind that platform. If they do not, list the blocked platforms and ask which one to rebind.

Do not change unrelated platform statuses except for a timestamped recheck if the user asks for a full health check.

生成今日 AI-HOTDog 热点报告

Run the configured radar:

1. Load the workspace configuration and source status.
2. If configuration is missing or incomplete, stop scanning and run 初始化 AI-HOTDog first.
3. Run the auth preflight gate from references/login-and-access.md.
4. Recheck stable public sources first.
5. Use only login-state platforms marked available or limited for discussion signals.
6. Collect candidates with title, URL, source id, source type, region, capture time, and matched keywords.
7. Score candidates for freshness, credibility, discussion, impact, and content value.
8. Deduplicate related reports into coherent hotspot clusters.
9. Generate 热点 Top 10, 自媒体选题 Top 5, one strongest recommendation, platform status, source-share stats, and a citation table.

Read references/report-template.md before writing the report. Run scripts/check_report_integrity.py on saved reports when a file is produced.

创建 AI-HOTDog 每日自动化

When the user asks for a recurring run, use the Codex automation tool if available. The automation prompt must describe only the task, not the schedule:

Use $ai-hotdog to load the existing AI-HOTDog configuration in this workspace, generate today's cited hotspot report, include source-share statistics, Top 10 hotspots, Top 5 content ideas, platform access status, and any platforms that need rebinding.

Only create the automation after .ai-hotdog/config.json exists. If selected login-state platforms have never been authorized, tell the user to run 授权登录 AI-HOTDog first or accept that the automation will run public sources only.

The automation job must not perform first-time login, QR-code scan, or credential entry. If a platform loses access during an unattended run, the report must list it under 需要重新绑定的平台 and continue with public sources.

Data Quality Gate

Before considering a report complete, confirm all of the following:

• 数据来源概览 includes scanned platforms, successful platforms, limited/blocked platforms, total candidates, scored candidates, final cited items, and source-share percentages.
• 热点 Top 10 contains exactly 10 ranked items unless the report explicitly says that strict citations left fewer than 10 valid items.
• 自媒体选题 Top 5 contains 5 ideas, each with a title, core point, platform fit, hook, structure, source material, and risk note.
• Each hotspot has a primary fact source, optional discussion sources, capture time, and confidence label.
• Every citation in 引用与来源表 maps to a visible hotspot, content idea, or source-status note.
• Any failed or limited source is disclosed in 平台访问状态 and never silently omitted from the source count.
• If selected login-state sources were skipped, 数据来源概览 must say the report is based on public sources plus any authorized platforms only.

Resource Routing

• Read references/source-registry.md during initialization, source customization, source-share calculation, or source-list debugging.
• Read references/config-schema.md before creating, editing, or troubleshooting workspace state.
• Read references/login-and-access.md before binding, rebinding, browser checks, or explaining account-access limits.
• Read references/report-template.md before generating or revising a report.
• Use scripts/init_ai_hotdog.py --workspace <workspace> ... to create deterministic non-secret workspace state.
• Use scripts/validate_source_registry.py <source-registry.md> after editing a source registry.
• Use scripts/check_report_integrity.py <report.md> before treating a saved report as finished.
• Use scripts/self_check.py before publishing the skill package.

Default AI Profile

If the user wants the default AI setup, initialize:

• focus_topic: AI, large models, AI tools, AI startups, AI content opportunities.
• regions: global and China.
• content_goal: content ideas.
• report_target: Top 10 AI hotspots and Top 5 content ideas.
• source_strategy: stable public sources first, login-state sources as supplemental discussion signals.
• strict_citation: true.