uzi

Warn

Audited by Socket on Sep 6, 2026

5 alerts found:

Anomalyx4Security
AnomalyLOW
run.py

No clear evidence of intentionally malicious payloads is present in this fragment. However, it performs security-sensitive supply-chain/bootstrap actions: (1) runtime pip installs (with --trusted-host derived from mirror URL) and (2) optional download/install of the cloudflared binary via curl/bash/sudo without shown integrity verification. It also modifies sys.path/cwd (increasing impact if packaged scripts are tampered with) and, in --remote mode, exposes local report content via a public tunnel URL. Treat this as a moderate supply-chain/operational security risk requiring dependency pinning, mirror hardening, and integrity verification for downloaded binaries.

Confidence: 60%Severity: 55%
AnomalyLOW
skills/deep-analysis/run.py

The code does not show overt malicious behavior, but it contains meaningful supply-chain and exposure risks: (1) it auto-installs missing Python dependencies via pip (potentially unpinned, executable code from package indexes), (2) it can download and install the cloudflared binary with sudo from GitHub via curl when `--install-cloudflared` is used, and (3) it deliberately exposes locally generated reports to the internet via Cloudflare Tunnel in `--remote` mode. Indirectly, loading internal modules from a modified sys.path and running an analysis pipeline means the real risk may depend on those imported modules, which are not included here.

Confidence: 62%Severity: 55%
AnomalyLOW
skills/deep-analysis/scripts/lib/versus_runner.py

This module is primarily a local report generator with no direct evidence of credential theft, network exfiltration, or embedded malware in the shown fragment. However, there is a meaningful local security risk: ticker-derived values are used in filesystem paths with insufficient sanitization. This can potentially enable path traversal for cache reads and for report output writes (write/read outside the intended directories) if an attacker can influence tickers (directly or indirectly through parse_ticker/loaded cached data). Additionally, the module executes an internal pipeline per ticker, which is a supply-chain/execution hotspot outside this fragment. Security review should therefore prioritize path handling and the behavior of the imported pipeline/module functions.

Confidence: 55%Severity: 58%
SecurityMEDIUM
INSTALL-HERMES.md

No direct malicious code is present in this fragment because it is installation/usage documentation. However, it instructs executing a remotely fetched installer script via `curl ... | bash`, and it sets up a chain of repo/dependency installation that can enable harmful behavior if any upstream component is compromised. Treat the installer and referenced skill code/dependencies as high-priority audit targets; do not rely on the scanner/bypass narrative alone.

Confidence: 48%Severity: 70%
AnomalyLOW
setup.sh

No explicit malicious payload is evident in this Bash installer fragment (no exfiltration, backdoor routines, or dynamic execution). However, it performs high-risk supply-chain operations: it fetches unverified repository code via `git clone/pull` and then installs dependencies from `requirements.txt` using pip across multiple mirrors, without hash/signature verification, pinning, or lockfiles. It also enables hook scripts via `chmod +x`, which could later be executed by the application. Overall, the dominant finding is elevated supply-chain compromise risk rather than confirmed malware in this specific file.

Confidence: 62%Severity: 66%
Audit Metadata
Analyzed At
Sep 6, 2026, 02:58 PM
Package URL
pkg:socket/skills-sh/wbh604%2Fuzi-skill%2Fuzi%2F@650788c54a9b2e042bf7f983f16dbf8d727fce1d
Security Audit — socket — uzi