SECURITYAuto-scan policy
Every listing, every push.
Auto-scan is the static + behavioral analysis layer that runs on every artifact before it goes live, and re-runs on every update. It is not optional. There is no "skip scan" flag. Sellers who try to obfuscate get banned, not warned.
CRITICAL
Wallet drainers, RCE, secret exfiltration. Auto-blocked, seller suspended pending review.
HIGH
Network access outside manifest, unsafe eval, license violation. Listing held; seller has 24h to fix.
MEDIUM
Outdated transitive deps, weak crypto, missing input validation. Listing goes live with a yellow badge.
LOW
Style nits, missing docstrings, lint warnings. Surfaced to the seller, hidden from buyers.
/03What buyers see
Every listing page carries a scan badge. Hovering reveals the date of the last scan, the ruleset version, and a link to the full report. Sellers can not edit or delete past scan reports — they are append-only and content-addressed.
/04Re-scan triggers
- Every listing update (any field, any artifact byte change).
- Weekly, on the listing's publish anniversary, with the latest ruleset.
- On any user-submitted abuse report, within 4 hours.
- On any upstream advisory (CVE, npm advisory, pip advisory) for declared dependencies.
/05Manual review
About 2.4% of scans escalate to manual review by a human on the trust & safety team. Median manual review time is 6 hours. The seller is notified and the listing is paused (not delisted) until the review concludes.