Token Estimator
PassAudited by ClawScan on May 1, 2026.
Overview
This skill appears to estimate token counts locally, with no credential use or data exfiltration shown; the main thing to notice is its reliance on third-party tokenizer packages and model files.
This looks safe for local token estimation. Before installing, be comfortable with adding the listed Python tokenizer dependencies and with the script loading external tokenizer resources for some models. Avoid pasting highly sensitive text unless you are comfortable processing it in this local environment and its installed libraries.
Findings (1)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The first use may rely on external tokenizer assets, and changes or compromise in those dependencies could affect local estimation behavior.
For Qwen/DashScope models, the skill uses Hugging Face Transformers to load an external tokenizer model. This is expected for token estimation, but it means behavior depends on third-party package/model provenance.
return AutoTokenizer.from_pretrained("Qwen/Qwen-7B")Install dependencies from trusted sources, consider pinning package/model versions for reproducibility, and expect tokenizer resources to be loaded for Qwen-style models.