Super Ocr
v0.1.0Production-grade OCR with intelligent engine selection. Tesseract (lightweight, fast) and PaddleOCR (high accuracy, Chinese-optimized). Use when extracting t...
⭐ 0· 293·4 current·4 all-time
byNima Chu@nimachu
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The registry metadata claims 'no required binaries / env vars' and 'instruction-only', but the shipped code clearly expects external binaries and packages: it checks for the tesseract binary, may invoke the Swift runtime on macOS, and depends on PaddleOCR/paddlepaddle and Python packages. The SKILL.md and README also advertise a library API (from super_ocr import OCRProcessor) but the repo appears to be a scripts/ tree with no packaging/setup to expose that module—this mismatch between advertised usage and actual files is inconsistent and may confuse installers.
Instruction Scope
Runtime instructions encourage running scripts/dependencies.py which can auto-install packages (pip installs) and run system commands. The SKILL.md content was flagged by a pre-scan for 'unicode-control-chars' (prompt-injection pattern), which is unexpected in documentation and could be used to interfere with downstream automated parsing or LLM-based evaluations. The instruction set does not ask for unrelated secrets, but it does instruct the agent/operator to execute install and runtime commands that affect the host environment.
Install Mechanism
There is no formal install spec in the registry, but the repository includes requirements.txt and a dependencies.py that runs 'pip install ...' as an auto-installer (subprocess.run). This is a common approach but means installing packages from PyPI (and PaddlePaddle's official wheel host). No suspicious custom download URLs or IP addresses were found; installs are via pip and standard package URLs (including a paddlepaddle wheel index). Still, auto-install behavior should be run in a controlled environment (virtualenv/container).
Credentials
The skill does not request environment variables or credentials. The code uses subprocess to invoke local binaries (tesseract, swift) and pip; it does not contain hardcoded external API keys or secret exfiltration endpoints. No config paths requiring broad system access are declared.
Persistence & Privilege
The skill is not marked always:true and does not request autonomous elevated privileges. It does not appear to modify other skills or system-wide agent configuration. It writes temporary processed images to the same directory when running OCR (e.g., _tess_processed.png) and removes them; this is normal for image-processing scripts.
Scan Findings in Context
[unicode-control-chars] unexpected: Detected in SKILL.md content/documentation. While presence of multilingual content and non-ASCII characters is expected for an OCR skill, unicode control characters in a skill manifest or SKILL.md are unexpected and were flagged as a prompt-injection pattern; review the raw SKILL.md for hidden characters or manipulative sequences before automated consumption.
What to consider before installing
This package looks like a real multi-engine OCR tool, but there are inconsistencies you should resolve before running it: (1) The registry metadata says 'no required binaries', yet the code expects the tesseract binary and (on macOS) the Swift runtime—ensure you actually want those installed. (2) The SKILL.md/README mention a Python library API but the repo is a scripts/ tree without an obvious package setup—don't assume 'from super_ocr import OCRProcessor' will work out of the box. (3) dependencies.py can auto-install packages with pip; run it inside a fresh virtual environment or container and inspect the install commands first. (4) The SKILL.md was flagged for unicode control characters—open the file in a hex/visible-control-char aware editor to ensure nothing suspicious is hidden. Recommended steps: run the code in an isolated VM/container, inspect dependencies.py and any subprocess commands, avoid running auto-install as root, and verify the swift/tesseract calls and temp-file behavior. If you need higher assurance, request the author to (a) provide a setup.py/pyproject for clean installation, (b) fix documentation filename/API mismatches, and (c) remove unexpected control characters from SKILL.md.Like a lobster shell, security has layers — review code before you run it.
latestvk97arpc3x3v88b65fk2k4mkpph82eac8
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
