LegalCorporate & Governance

Corporate Registry Investigation

Use when users say "Companies House search", "investigate this UK company", "check officers/PSCs/charges", "registry snapshot", "group structure", "filing history", or need UK company profile, filings…

LLegalQuants·Legal·Apache-2.0

Library skill — the default version is maintained in GitHub; edits you make live in your own clone.

Use this skillDownload .zip
How does this work?
  • ChatGPT opens a new chat with the skill loaded. If it's too long for a link, it's copied to your clipboard — just paste.
  • Claude works the same way. To install it permanently, download the .zip and upload it under Claude → Settings → Capabilities → Skills (Pro/Team/Enterprise).
  • Copy prompt copies the skill so you can paste it into any assistant, including Grok.

corporate-registry-investigation

When to Use

This skill reads registry evidence. It does not prove beneficial ownership beyond what the registry and supplied documents show.

Audience and Work Shape

Audience: disputes, diligence, KYC, onboarding, conflicts, and in-house legal users doing first-pass UK registry review for lawyer or compliance review.

Work shape: pattern-matched review of registry evidence. The skill extracts and labels public registry facts and risk leads; it does not conclude wrongdoing, beneficial ownership, or suitability.

Legal Failure Modes

Access Modes

This skill works in three modes:

  1. Live registry mode - use browser, web search, MCP, API, or other configured access to retrieve Companies House pages, filings, officers, PSCs, charges, and filing history.
  2. User-supplied registry mode - use company profile pages, filing PDFs, screenshots, API exports, CSVs, HTML, or pasted registry material supplied by the user.
  3. No-source mode - prepare a company-number/name search plan and checklist, but do not state registry facts or risk signals.

If live Companies House access is unavailable and the user has not supplied registry material, ask for a company number, company-profile export, filing PDFs, screenshots, or source access. Do not fill registry facts from model memory.

How It Works

1. Identify the company

Confirm:

Prefer company number over name matching.

2. Pull registry snapshot

If live Companies House access is unavailable, ask the user for filings, screenshots, an API export, or company-profile text. State clearly whether findings are based on live registry access or user-supplied registry material.

Collect:

Record retrieval date. Registry data changes. For each material fact, cite the filing name, filing date, transaction id, document id, or registry page from which it came.

3. Review people and control

Extract:

Do not conflate individuals across companies by name alone. Use available identifiers such as month/year of birth, correspondence address, appointment dates, officer id, and source filing; if identity is uncertain, mark possible_match_needs_verification.

Flag evidence-based registry patterns requiring verification, such as rapid officer churn, repeated service-address or corporate-officer patterns, circular control, or missing PSC statements. Do not infer wrongdoing, nominee status, or control from nationality, residence, or name patterns alone.

4. Review charges and filings

Check:

When available, read filing text rather than relying only on filing titles.

5. Build relationship and risk summary

Produce:

Label red flags as investigative leads, not legal findings or allegations.

Confidence Bands

Not in Scope

Escalation

Stop and ask the responsible lawyer or reviewer when identity match is uncertain, filings are unreadable or scanned without OCR, ownership cannot be resolved from registry material, or a pattern would need to be characterised as wrongdoing to be useful.

6. Keep evidence and uncertainty separate

Use:

Do not overstate public registry data.

Example

Investigate this UK company. Pull Companies House details, officers, PSCs, charges, filing history, accounts signals, and any red flags. Separate registry facts from inferences.

For a compact output pattern, see examples/output.md. For the registry snapshot, filing extraction, risk-lead, and identity-caution model, see references/companies-house-investigation-model.md.

Limitations