AI Business OSINT

Use this skill to run a public-source business investigation on AI training/consulting/event operators and convert the findings into an actionable business playbook.

Operating Boundaries

Use public, lawful, attributable sources: official company registries, government pages, procurement pages, universities, associations, enterprise sites, course platforms, public social profiles, and media reports.
Do not collect or publish private phone numbers, ID numbers, private WeChat IDs, non-public group QR codes, home addresses, cookies, or account/session secrets.
Separate verified facts, source-supported inferences, and open questions.
Browse current sources for laws, regulations, procurement rules, prices, dates, corporate data, and public positions.
Prefer primary sources; use aggregators only as leads unless the user explicitly provides a logged-in business database page.
When using logged-in Chrome or other user sessions, keep work read-only unless the user explicitly requests an action.

Workflow

Scope the target
- List target names, aliases, Chinese/English names, brands, company names, account handles, course names, and likely cities.
- Define the user goal: pure OSINT, competitor mapping, company deep dive, cooperation timeline, or personal entry strategy.
- Set exclusion rules for同名噪声 and private data.
Build the source map
- Search official/government sources first for companies, licenses, procurement, events, and policy rules.
- Search universities, associations, enterprise sites, course platforms, and media for public activities.
- Search public social platforms only for stable account pages, bios, article/video titles, and business-facing contact surfaces.
- Use business databases such as 企查查/天眼查 only to verify corporate closure: legal representative, shareholders, registered capital, paid-in capital, address, business scope, risk, trademarks, tenders, changes, and related entities.
Close identity loops
- Treat a company as belonging to a person only when at least one strong bridge exists: legal representative, shareholder, executive role, official company page, verified profile, repeated official event bio, trademark/brand linkage, or signed procurement result.
- Do not merge same-name companies into the target without a bridge.
- Mark conflicting sources explicitly and avoid overclaiming control.
Extract business model
- Identify whether the operator is company/IP-driven, platform/mentor-driven, institution-backed, content/community-driven, product/SaaS-driven, or procurement-driven.
- Capture product lines: executive briefings, enterprise training, public courses, camps, competitions, AI agent implementation, consulting retainers, certification programs, or community memberships.
- Record channels: universities, business schools, government departments, industry associations, unions, parks/incubators, state-owned enterprises, private enterprises, public social accounts, and private communities when publicly disclosed.
Create timelines
- Build chronological tables for company changes, trademarks, public activities, courses, procurement/tenders, government cooperation, enterprise cooperation, and major media appearances.
- Include exact dates when available; use month/year only when exact dates are not public.
- Distinguish activity reports from signed contracts and procurement awards.
Assess compliance
- Classify each activity: ordinary enterprise consulting/training, vocational training, certificate issuance, government procurement, large public event, livestream/online course, AI product/platform, advertising/claims, data handling.
- For each class, check current official rules and write practical implications.
- Avoid legal conclusions beyond available sources; recommend local lawyer/accountant confirmation for execution.
Turn findings into strategy
- Derive reusable tactics from the targets: positioning, entity setup, product packaging, channel strategy, proof assets, procurement readiness, pricing logic, and risk controls.
- For a from-zero user, produce a concrete sequence from company registration through first pilots, supplier onboarding, partnerships, and scale-up.

Deliverables

Choose the format that matches the request:

Public OSINT brief: identity, accounts, courses, events, social footprint, confidence levels.
Company deep dive: corporate entities, shareholders, legal representative, capital, scope, risks, IP, tenders, related companies.
Cooperation timeline: government/enterprise/university/association activities and contracts by date.
Market-entry playbook: company registration, business scope, tax/accounting, qualifications, product packages, go-to-market, procurement readiness, documents, risks.
Evidence matrix: claim, source, source type, confidence, next verification step.

Load references/research-checklist.md when producing a full report or playbook.

Report Standards

Put the strongest conclusion first, then evidence and caveats.
Use tables for entity data, timelines, product packages, compliance decision trees, and action plans.
Include source links, especially for official rules and public activity pages.
Use confidence labels when information is incomplete: high, medium, low, unverified.
Call out “not found” findings when meaningful, such as no closed-loop company for a public trainer.
Keep private-data boundaries visible in the report.