2026-W38

Published 2026-09-18 · static snapshot, no dynamic audit

← All reports
15,443
scanned
0
critical
0
high
108
medium
15,113
low

No dynamic audit in this archive

This archive predates the merge of the dynamic audit into weekly reports and records the static scan snapshot of that time (three tiers: low / medium / critical; the "high" tier was folded into medium). Published archives are frozen — today's dynamic conclusions are not backfilled into them. See the latest weekly report and each plugin page for the current caliber.

This week

Against the previous archive, in effective-tier terms. If the previous archive predates the dynamic merge, this diff mixes a caliber change with real movement.

▲ New medium · 2

Critical-risk plugins

0 plugins with an effective critical tier (static snapshot).

PluginTier
None

High-risk plugins

0 plugins with an effective high tier (static snapshot).

PluginTier
None

Medium-risk plugins

108 plugins with an effective medium tier (static snapshot).

PluginTier
dsh-remote★ 47medium
dsh-web-profile★ 0medium
dsh-smart-reminder★ 0medium
dsh-upgrade-kit★ 0medium
dsh-tmwebdriver★ 0medium
dsh-v-explorer★ 2medium
dsh-remote★ 0medium
claude-in-dsh★ 3medium
cli★ 7medium
dsh-remote★ 1medium
dsh-cordis-mcp★ 1medium
vision-toolkit-for-dsh-v0.1-maybe-★ 1medium
dsh-desktop★ 2medium
dsh-auto-classifier★ 1medium
dsh-launcher★ 0medium
dsh-balance-plugin★ 62medium
dsh-convmap★ 1medium
dsh-moodball★ 0medium
paper-workbench★ 0medium
dsh-balance★ 2medium

Showing the 20 with the most static findings; the remaining 88 are in the plugin list filters.

Weekly report archive

12 reports

What does this weekly report measure?

This report measures the registry's current risk state during 2026-W38: how many plugins were scanned, the distribution across four tiers (low, medium, high, critical), and which plugins changed tier since the previous archive. The tier is not a single static rule: where the dynamic audit reached a conclusion, that conclusion is authoritative; the static scan tier applies only where there is none.

How the dynamic audit relates to the static scan

The static scan is deterministic rule matching over source shape, manifests and permission requests, and covers the whole registry. The dynamic audit has a model read redacted source context (plugin code is never executed) for the subset whose static tier is medium or above, and judges context — what actually loads at runtime, and whether a list is defensive or offensive. Merge rule: a dynamic conclusion wins (it may raise or lower the tier); otherwise the static tier applies. If human review finds the model wrong, recording `withdrawn` in the review table reverts the tier to the static one immediately. Exemptions are manual only; the audit script never writes the suppression list.

Why isn't this report an endorsement?

Every row comes from public repository metadata, heuristic rule matches, or a model reading source context — all of which can be wrong: rules produce false positives, and a model can be fooled by injected source comments or simply misjudge. Risk flags are therefore a reason to inspect a plugin before install, not a substitute for that inspection.

Source: dsh.so security standards · Maintained by dsh.so · Published

Was this page helpful?