Agentic CV Repro Lab
v1.7.2Public ClawHub skill for execution-grade CV experiments and evidence capture across Colab, Kaggle, browser automation, and GPU VMs.
⭐ 0· 118·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description, SKILL.md, and the bundled Python scripts align: they create run cards, manifests, capture environment and GPU state, and provide browser/Colab/Kaggle guidance. Required binaries (python3/python) and the presence of git/nvidia-smi usage in scripts are appropriate for the stated purpose.
Instruction Scope
SKILL.md instructs the agent to initialize manifests, capture run context, and use the bundled scripts. The scripts run local commands (git, nvidia-smi) and inspect the environment/package metadata — actions coherent with 'capture experiment context'. I found no instructions to read unrelated host secrets, sweep arbitrary system files, or post data to unknown external endpoints. The code also explicitly sanitizes URLs, paths, and potential secrets before emitting durable artifacts.
Install Mechanism
This is an instruction-only skill with bundled helper scripts and no install spec. Nothing is downloaded or extracted at install time; risk from install mechanism is minimal.
Credentials
The skill requests no environment variables or credentials. The scripts invoke local tools (git, nvidia-smi) and inspect installed Python packages — reasonable for reproducibility capture. The included cv_public_safety sanitization functions deliberately redact sensitive-looking keys and credential-style URLs.
Persistence & Privilege
always is false; the skill is user-invocable and does not demand permanent presence or modify other skills. It writes run artifacts to paths supplied by the user (out flags) and uses path sanitization before emitting durable records.
Assessment
This skill appears coherent and appropriate for collecting reproducibility artifacts for CV runs. Before running: review the scripts locally (they will call git and nvidia-smi and write files where you point --out), ensure you execute it in a repository or workspace you control (don’t point it at directories containing secrets), and prefer running in a sandboxed environment if you are unsure. The code includes sanitization to avoid leaking absolute paths or credential-like values into public artifacts, but you should still verify outputs before publishing them externally.Like a lobster shell, security has layers — review code before you run it.
agentic-workflowsvk977wrr7hzs3byhj6sx7srxk5983a8zncolabvk977wrr7hzs3byhj6sx7srxk5983a8zncomputer-visionvk977wrr7hzs3byhj6sx7srxk5983a8zndata-sciencevk977wrr7hzs3byhj6sx7srxk5983a8zngpu-trainingvk977wrr7hzs3byhj6sx7srxk5983a8znkagglevk977wrr7hzs3byhj6sx7srxk5983a8znlatestvk977wrr7hzs3byhj6sx7srxk5983a8znmlopsvk977wrr7hzs3byhj6sx7srxk5983a8znopenclawvk977wrr7hzs3byhj6sx7srxk5983a8znreproducibilityvk977wrr7hzs3byhj6sx7srxk5983a8zn
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
Any binpython3, python