Korean Document Reviewer
ReviewAudited by ClawScan on May 1, 2026.
Overview
This instruction-only skill is coherent for Korean business document review, but users should notice that it handles sensitive financial/business data, may send it to subagents, and saves extracted results in the workspace.
Before installing, confirm that the documents you will review may be processed by the agent and any spawned Sonnet subagents. Use a trusted workspace, avoid uploading documents that should not be persisted, and clean up saved review JSON files or generated images when the review is complete.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Bank-account, tax, contract, and business-registration details may be handled by multiple spawned agent sessions during review.
The skill explicitly delegates document OCR/review to spawned Sonnet subagents and gives them file paths to read. This is disclosed and purpose-aligned, but it means sensitive business documents may be processed outside the main agent session.
각 그룹을 `sessions_spawn`으로 Sonnet 서브에이전트에 위임한다 ... 이미지 파일 경로를 명시하고 "Read tool로 읽어라"고 지시
Use this workflow only when subagent/model processing of the documents is acceptable; for highly confidential documents, ask for single-session review.
Extracted account numbers, business registration numbers, names, dates, and amounts may remain in workspace files after the chat summary is delivered.
The skill extracts sensitive business and bank fields and instructs saving the structured result in the workspace. This is useful for the stated purpose, but it creates persistent copies of sensitive document data.
핵심 필드: 사업자등록번호, 상호, 대표자, 금액, 날짜, 계좌번호 ... JSON 결과를 workspace에 저장: `reviews/REV-{YYYYMMDD}-{NNN}.json`Store review outputs only in trusted workspaces, control sharing of the reviews folder, and delete or redact saved JSON files when no longer needed.
PDF files supplied for review may be converted into image files in the local environment.
The skill uses a local PDF conversion command as part of OCR preprocessing. This is central to the document-review purpose and is not hidden, but users should recognize that local files are processed by an external binary.
PDF를 이미지로 변환 (`pdftoppm -png -r 200`)
Run the skill in a trusted workspace with a known-good PDF conversion tool, and review or clean up generated images if needed.