AI Workstation Topic Intelligence

Find current topics worth researching, then turn one into a clear, actionable content brief.

简体中文

Topic Intelligence is an installable Agent Skill. It uses current public signals from the AI Workstation Global Topic Radar to help creators, researchers, and editors:

• shortlist topics worth deeper research;
• turn a supplied current topic into a research-ready brief;
• select one topic and continue directly into a brief for that same topic.

The Radar supplies current signal metadata and source links. Your Codex or compatible Agent Skills host performs the analysis; material external claims still require primary-source verification.

Version 0.4 aligns the Skill with the Radar's editorial lanes (curated, rising, market_lead, early, watch, evergreen, raw, and all), language/market-aware ranking, four component scores, evidence states, and independent-publisher provenance. The Skill chooses these query dimensions explicitly instead of inheriting API defaults.

Get started

1. Open the latest GitHub Release for the standalone Skill archive, or clone this repository for the Codex installer.
2. Install the Skill using the host-specific workflow below.
3. Describe whether you want topic selection, a brief, or both.

Try:

Choose one current AI topic for a two-to-three-minute explainer, then turn it into a research-ready content brief. Include the audience payoff, strongest angle, opening, narrative structure, must_verify, avoid_claims, and suggested visuals.

Installation

Codex

python3 scripts/install_codex_skills.py install
python3 scripts/install_codex_skills.py doctor

Other Agent Skills hosts

Import the Release ZIP or topic-intelligence directory. The host must be able to read SKILL.md, run the bundled Python helper, and access the public Radar API.

See the installation guide for detailed steps and first-use prompts.

Product pages: English · 中文

What the brief includes

• audience payoff and editorial angle;
• opening and narrative structure;
• research questions and source priorities;
• must_verify and avoid_claims;
• visual and material requirements.

Evidence and privacy

• Current claims must come from a current Radar response, not model memory or saved snapshots.
• Radar observations and evidence links are research leads, not independent verification of external claims.
• Radar's editorial, trend, confidence, and market-fit scores are deterministic estimates, not a fifth host score or a performance forecast.
• Evidence breadth is interpreted by independent publisher identity, not by the number of collectors displaying the same origin.
• Incomplete or older source coverage must be disclosed.
• Radar does not measure actual audience size, content saturation, real demand, post-publication performance, or future reach/virality.
• The public Skill uses read-only Radar endpoints and requires no AI Workstation API key.
• Do not include credentials, private conversations, or client data in public issues.

Development

python3 scripts/sync_skill_runtime.py --check
python3 scripts/sync_plugin_candidate.py --check
python3 -m unittest discover -s tests -v
python3 -m compileall -q scripts skills

See CONTRIBUTING.md for contribution requirements and GitHub Issues for support and feedback.

License

Apache-2.0. See LICENSE.