run-corpus-research

SUSPICIOUS: the skill’s stated purpose and direct behavior are largely coherent for research orchestration, and it shows no direct credential theft, exfiltration endpoint, or installer abuse. However, it embeds mandatory use of downstream skills/tools with unspecified provenance and relies on web-derived subagent outputs, creating transitive-trust and indirect prompt-injection risk that is disproportionate to a purely local orchestration skill unless those dependencies are separately verified.