← Back to Skills Marketplace
415
Downloads
2
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install qualia-skill
Description
Fine-tune robot foundation models on cloud GPUs — π0.5, π0, GR00T, SmolVLA, ACT, and more.
Usage Guidance
This skill appears internally consistent and implements a CLI for the Qualia service. Before installing: (1) Confirm you trust https://qualiastudios.dev and the api.qualiastudios.dev endpoint; (2) Only provide a Qualia API key with the minimum necessary scope/credits and be aware training costs (scripts call instances and estimate credits/hr); (3) If you use private datasets, note the skill will reference HuggingFace dataset IDs and send them to the Qualia API—ensure that aligns with your data privacy needs; (4) The skill runs bundled shell/Python scripts that will perform network requests — run them in a controlled environment or inspect them locally if you have concerns; (5) Revoke or rotate the API key if you stop using the skill. Overall this looks coherent for its stated purpose.
Capability Analysis
Type: OpenClaw Skill
Name: qualia-skill
Version: 2.0.0
The Qualia skill bundle is a legitimate toolset for interacting with the Qualia robotics fine-tuning platform. The provided scripts (qualia.py and qualia.sh) are clean API wrappers using standard libraries to manage training jobs, projects, and credits via https://api.qualiastudios.dev. No evidence of data exfiltration, malicious execution, or prompt injection was found.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description, SKILL.md, README, and the two CLI scripts all focus on interacting with api.qualiastudios.dev to list models, inspect datasets, create projects, and launch/monitor finetune jobs; the only required environment variable is QUALIA_API_KEY which directly matches the described API access.
Instruction Scope
Runtime instructions and the scripts only reference HuggingFace dataset IDs, model/instance selection, camera mapping, and calls to the Qualia API. They do not instruct reading unrelated local files, other environment variables, or sending data to unexpected external endpoints.
Install Mechanism
No install spec is present (instruction-only with included scripts). The bundled bash and python scripts are self‑contained and make HTTPS requests to the documented API host; there are no downloads from third‑party URLs or archive extraction steps.
Credentials
Only one env var is required (QUALIA_API_KEY). That is proportionate to the task; no broad or unrelated credentials/config paths are requested. The code accesses no other environment variables.
Persistence & Privilege
The skill is not force‑included (always: false) and does not modify other skills or system configuration. It relies on user-supplied API credentials and user invocation, so its system presence and privileges are minimal and expected.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install qualia-skill - After installation, invoke the skill by name or use
/qualia-skill - Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.0.0
Changes in v2:
New: scripts/qualia.py — Complete Python rewrite of the CLI. Pure stdlib (urllib + json), zero external dependencies. No more curl/jq requirement. Same commands, same interface. This is what was triggering VirusTotal (bash + curl + API keys = "suspicious data exfiltration" heuristic).
Updated: SKILL.md — Better agent instructions: onboarding flow with sign-up link, "when someone asks to train" decision tree so agents know what info to gather, troubleshooting table, smart camera mapping hints, cost estimation guidance, live docs links. Added ClawhHub metadata: tags (robotics, foundation-models, vla, etc.), categories, and homepage (qualiastudios.dev).
Updated: README.md — References changed from bash to Python, simplified requirements to just "Python 3.6+".
Kept: scripts/qualia.sh — Still there for backwards compatibility. Not referenced in SKILL.md anymore but won't break existing installs.
v1.0.0
Initial release — fine-tune and iterate on robotic foundation models on cloud GPUs
Metadata
Frequently Asked Questions
What is Train Robotic AI Models using Qualia?
Fine-tune robot foundation models on cloud GPUs — π0.5, π0, GR00T, SmolVLA, ACT, and more. It is an AI Agent Skill for Claude Code / OpenClaw, with 415 downloads so far.
How do I install Train Robotic AI Models using Qualia?
Run "/install qualia-skill" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Train Robotic AI Models using Qualia free?
Yes, Train Robotic AI Models using Qualia is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Train Robotic AI Models using Qualia support?
Train Robotic AI Models using Qualia is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Train Robotic AI Models using Qualia?
It is built and maintained by Qualia (@fabbe1999); the current version is v2.0.0.
More Skills