bluebook-audit

SUSPICIOUS: the skill's main verification purpose is benign, but it includes a non-essential load-time pre-execution command that runs an external local helper binary from PATH or the user's home directory. There is no clear credential theft or exfiltration, so this is not confirmed malware, but the execution model is broader and riskier than necessary for a DOCX verification skill.

SUSPICIOUS. The stated purpose is a benign legal citation audit, but the skill includes load-time pre-execution of an unresolved local helper binary (`skill-toc`/plugin-utils). That capability is unnecessary for Bluebook checking, introduces pre-review execution risk, and triggers the unverifiable-executable floor even without confirmed exfiltration or malware.

SUSPICIOUS. The core DOCX-correction purpose is coherent, but the skill uses load-time pre-execution and delegates to an externally installed helper of unclear provenance, plus it chains into another skill. I found no direct credential theft, exfiltration, or remote-code fetch here, so this is not confirmed malware; the main issue is disproportionate trust in unverifiable local plugin code executed before normal oversight.

SUSPICIOUS. The skill’s stated reporting purpose is coherent and there is no visible exfiltration, credential access, or remote download. However, it contains a load-time pre-execution command that runs an unverifiable local plugin binary (skill-toc/plugin-utils) with unclear provenance, which is disproportionate to a report-writing skill and raises supply-chain trust concerns.

SUSPICIOUS. The main DOCX extraction workflow is coherent and same-repo aligned, with no obvious credential theft or off-platform data flow. However, the skill contains a load-time pre-execution directive that may run an unverified local plugin binary before the user or agent can inspect the content, which is disproportionate to a documentation/parsing step and creates medium supply-chain and execution-trust risk.