← Back to Skills Marketplace
Perceptron
by
Subraiz Ahmed
· GitHub ↗
· v1.0.1
301
Downloads
2
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install perceptron
Description
Image and video analysis powered by Isaac vision models. Capabilities include visual Q&A, object detection, OCR, captioning, counting, and grounded spatial r...
Usage Guidance
This package appears coherent for image/video analysis, but before installing: 1) Understand that images/frames you analyze will be sent to the Perceptron API (api.perceptron.inc is referenced in docs); do not upload sensitive images unless you trust the provider. 2) The skill requires an API key (PERCEPTRON_API_KEY) — grant only a key with the minimum needed scope and rotate it if exposed. 3) The README instructs pip install perceptron — verify the perceptron package on PyPI and the official docs (https://docs.perceptron.inc/) to ensure you're installing the intended SDK. 4) Batch/CLI operations read/write local files and can process directories; be careful when running with paths that contain sensitive files. 5) If you allow the agent to invoke skills autonomously, remember it can use the provided API key to call the service; restrict autonomous use if that is a concern.
Capability Analysis
Type: OpenClaw Skill
Name: perceptron
Version: 1.0.1
The 'perceptron' skill bundle provides a legitimate interface for image and video analysis using the Perceptron SDK and Isaac vision models. The core logic in `scripts/perceptron_cli.py` functions as a standard CLI wrapper for tasks like object detection, OCR, and visual Q&A, communicating with the official API endpoint at https://api.perceptron.inc. No evidence of data exfiltration, unauthorized execution, or malicious prompt injection was found; the code follows the stated purpose and uses standard security practices like masking API keys in configuration displays.
Capability Assessment
Purpose & Capability
Name/description, the included CLI, SKILL.md, and the single required env var (PERCEPTRON_API_KEY) all align with a vision-analysis SDK that calls an external Perceptron service.
Instruction Scope
Runtime instructions and the CLI operate only on image/video files, use the Perceptron SDK, and optionally read/write local files for batch processing or output. The documentation does not instruct reading unrelated system files or credentials. Note: using the skill will upload images/frames to the Perceptron service, so any sensitive image data will be transmitted externally.
Install Mechanism
There is no registry install spec; SKILL.md instructs pip install perceptron (standard). The skill includes a Python CLI that depends on that SDK and optional Pillow. No downloads from untrusted URLs or archive extraction are present.
Credentials
Only PERCEPTRON_API_KEY (primary credential) is required, which is appropriate for an external API client. No unrelated credentials or config paths are requested.
Persistence & Privilege
always is false and the skill is user-invocable; it does not request persistent system-wide privileges or modify other skills. Note that autonomous invocation (the platform default) would allow the agent to call the Perceptron API using the provided key, which is expected behavior but something to be aware of.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install perceptron - After installation, invoke the skill by name or use
/perceptron - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.1
- Simplified and clarified the description for general audiences; removed strict usage enforcement and emphasized capabilities.
- Added structured metadata for required environment variables.
- Focused documentation on capability overview, setup, SDK usage, and CLI, omitting mandatory language about routing all vision through Perceptron.
- Removed internal enforcement instructions and narrowed guidance to user-facing APIs and practical usage.
- No code or file-level changes; documentation and presentation improvements only.
v1.0.0
- Initial release of the perceptron skill for image and video analysis.
- Provides mandatory routing of all visual content (images, videos, screenshots, OCR, object detection, etc.) through Perceptron — native multimodal vision is disabled.
- Includes Python SDK functions for Q&A, object detection, OCR, captioning, counting, and advanced DSL workflows.
- Offers CLI script for batch and interactive visual analysis tasks.
- Supports structured outputs, pixel coordinate conversion, model selection, and detailed documentation references.
Metadata
Frequently Asked Questions
What is Perceptron?
Image and video analysis powered by Isaac vision models. Capabilities include visual Q&A, object detection, OCR, captioning, counting, and grounded spatial r... It is an AI Agent Skill for Claude Code / OpenClaw, with 301 downloads so far.
How do I install Perceptron?
Run "/install perceptron" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Perceptron free?
Yes, Perceptron is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Perceptron support?
Perceptron is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Perceptron?
It is built and maintained by Subraiz Ahmed (@subraiz); the current version is v1.0.1.
More Skills