Back to skill
Skillv1.0.26

ClawScan security

WiseDiag MedOCR · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 31, 2026, 7:28 AM
Verdict
Benign
Confidence
high
Model
gpt-5-mini
Summary
The skill's code and runtime instructions match its stated purpose (upload a single PDF/image to WiseDiag for OCR and save Markdown locally); no unexpected credentials, endpoints, or install sources are present, though the registry metadata omits the required API key declaration.
Guidance
This skill appears to do what it says: it uploads one file to WiseDiag's cloud OCR API and saves the returned Markdown locally. Before installing or using it, consider: 1) Do NOT upload sensitive documents — the SKILL.md explicitly warns against IDs, financial, medical, biometric, credentials, or minor data. 2) The script requires WISEDIAG_API_KEY; the package metadata omitted this, so be sure to set the environment variable rather than assume none are needed. 3) The script calls https://openapi.wisediag.com; verify that domain and WiseDiag's privacy/data-retention policy meet your requirements. 4) If you must process sensitive files, use an offline OCR alternative. 5) As a best practice, set the API key in a secure location (avoid committing it to dotfiles or repos). Finally, if you want extra assurance, review network traffic or run the script in an isolated environment (VM/container) while testing with non-sensitive files.

Review Dimensions

Purpose & Capability
noteThe skill name/description (WiseDiag OCR to Markdown) aligns with its behavior: the script uploads a single PDF/image to WiseDiag's OCR endpoint, receives a JSON result, and writes Markdown to ~/.openclaw/workspace/WiseOCR. Note: the top-level registry metadata in the package summary listed no required env vars, but the SKILL.md and the script both require WISEDIAG_API_KEY — this metadata omission is an inconsistency to be aware of.
Instruction Scope
okSKILL.md instructs the user to set WISEDIAG_API_KEY and run the provided Python script. The script only reads the input file, optionally inspects PDF page count, uploads the file to https://openapi.wisediag.com/v1/ocr/pdf, and saves returned markdown locally. It does not access other system files, other credentials, or unusual endpoints. The privacy warning in SKILL.md accurately reflects the upload behavior.
Install Mechanism
okThere is no custom install step; the project includes a requirements.txt (requests, pypdf) and recommends pip install -r requirements.txt. No remote arbitrary installers, URL downloads, or packaged binaries are present. The code file is plain Python and straightforward to inspect.
Credentials
noteThe only runtime secret required is WISEDIAG_API_KEY, which is reasonable and proportional for a cloud OCR service. However, the registry-level 'Requirements' field (provided earlier in the package metadata) did not list this env var while SKILL.md and the script both require it — this mismatch could confuse users or automated tooling.
Persistence & Privilege
okThe skill runs on demand (always:false) and is user-invocable. It writes output only to a subdirectory under the user's home (~/.openclaw/workspace/WiseOCR). It does not modify other skills, system-wide settings, or request permanent elevated privileges.