Skillv1.0.0

ClawScan security

A professional skill for reading electronic datasheets · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 13, 2026, 1:52 AM

Verdict: benign
Confidence: medium
Model: gpt-5-mini
Summary: The skill's files, instructions, and included PDF tool align with its stated purpose (extracting datasheet info from PDFs); it requests no credentials or unusual system privileges, though it includes a local Python tool and requires PDF parsing/rendering dependencies that would need to be installed at runtime.
Guidance: This skill is internally coherent and appears to implement what it claims: a PDF-centric datasheet extractor backed by a local Python tool. Things to consider before installing: (1) source provenance — there is no homepage and the publisher identity is just an owner ID, so review scripts/pdf_tools.py yourself if you require higher trust; (2) runtime dependencies — using render_page requires installing pdf rendering Python packages (pymupdf or pypdfium2) via pip, which your environment may need to allow; (3) runtime behavior — the tool reads user-supplied PDFs and may write temporary files (pattern JSON, rendered PNGs); run it in a sandbox if the PDFs are sensitive; (4) performance — page_hints/scan over large PDFs can be slow and CPU/memory intensive; and (5) the SKILL.md strongly instructs the agent to avoid guessing and to source values only from the PDF — that is a policy choice but not a security risk. If you want higher confidence, ask the publisher for a homepage or signed release, or inspect/execute the python script in a controlled environment first.

Review Dimensions

Purpose & Capability: okThe name/description match the deliverables: SKILL.md describes a PDF-first extraction workflow and the repository contains templates and a pdf_tools.py script implementing search, table extraction, rendering, and page heuristics. There are no unrelated environment variables, binaries, or secrets requested that would be inconsistent with a datasheet-extraction skill.
Instruction Scope: okRuntime instructions constrain the agent to extract only from the PDF (Iron Law) and provide a stepwise 6-phase workflow using the included scripts. The instructions reference reading and writing temporary files (e.g., /tmp patterns JSON) and rendering pages to images for image-only PDFs, which is appropriate for this purpose. There are no instructions to read unrelated system files, environment secrets, or to exfiltrate data to unknown endpoints; the guidance to 'search the manufacturer website' is a manual follow-up suggestion, not an automated network action in the SKILL.md.
Install Mechanism: noteThere is no formal install spec in registry metadata. The README documents Python dependencies (pdfplumber plus either pymupdf or pypdfium2) which would need to be installed via pip to use render/image commands. This is expected for a local PDF-processing tool but means the runtime environment may install Python packages; there are no downloads from untrusted URLs or archive extraction steps in the skill files themselves.
Credentials: okThe skill declares no required environment variables, credentials, or config paths. The code does not request secrets and the templates and scripts operate on user-supplied PDF paths. No disproportionate access to cloud or system credentials is requested.
Persistence & Privilege: okThe skill is not marked always:true and is user-invocable (normal defaults). It does not request or attempt to modify other skills or system-wide agent settings. Autonomous model invocation is allowed (disable-model-invocation:false) which is the platform default and not itself a negative signal here.