AI risk level

Low

Recommendation

install

Model

~deepseek/deepseek-v4-flash-latest

Analyzed commit

a8078fa

Reviewed

5 days ago

The deterministic scan found no issues, and I agree there is no obviously dangerous behavior: the plugin is readable, uses no shell injection, has no install hooks, and only runs a Python collector on a timer. I rate it low rather than none because it reads Copilot OAuth tokens from editor configs and the gh CLI and makes an authenticated network request to GitHub, but this is transparent, matches the stated purpose, and the sampled code sends the token only to api.github.com.

• • The collector reads OAuth tokens from ~/.config/github-copilot (apps.json, hosts.json, oauth.json) and via `gh auth token`; this is sensitive, though the sampled code uses them only for the GitHub quota API and does not appear to persist or exfiltrate them.
• • The plugin runs a Python process every 5 minutes and keeps itself loaded, but the operations are limited to reading local Copilot data, fetching quota, and writing a usage record under the user's state directory.

How this check works

This review combines the deterministic scan (the rule-based results above) with an independent look at the plugin's code by a language model. The model reads a trimmed sample of the repository's files, the manifest, and the README, then gives a plain-language risk level and a recommendation: install (no notable danger), review (look closer first), or avoid (clearly dangerous).

It runs on the same analyzed commit as the deterministic scan and is strictly advisory — it is not a security guarantee and never blocks a plugin by itself. A human moderator still reviews plugins before they are listed.

AI advisory only — automated analysis, not a security guarantee.