leo-live-inspector

SUSPICIOUS. The skill's purpose broadly matches internal SRE/ops workflows, and data mainly flows to official internal endpoints rather than attacker infrastructure. But its footprint is disproportionately broad and high-impact: it harvests/persists multiple session tokens, silently runs backend scripts, reads local env/config secrets, and enables test-environment writes and message sends with incomplete confirmation safeguards. High security risk from autonomy and credential concentration, but not confirmed malware.

The code does not show overt malware (no external exfiltration, no reverse shell/process execution), but it is highly security-sensitive. It can harvest MySQL credentials from local .env/application config files when not provided, then optionally persists those credentials (including passwords) to a local JSON cache, and it executes arbitrary SQL from the command line directly against a MySQL server with only heuristic safeguards (and an explicit bypass via --force). In a supply-chain context, this increases the impact of running the tool in the wrong context or with untrusted inputs, even if the intent appears to be developer/test tooling.

The code implements a remotely controlled browser automation extension with highly sensitive capabilities. It polls a configurable gateway/bridge over HTTP for tasks, can execute attacker-supplied JavaScript in the page via window.eval(params.script), and can read cookies and exfiltrate them back to external endpoints via fetch. While such functionality could be legitimate for testing/automation, the presence of a remote code execution primitive and direct cookie exfiltration makes the module strongly suspicious from a supply-chain/malware standpoint.

The code is an administrative CLI that submits user-selected SQL to a hardcoded internal cloud MySQL service, retrieves an XLSX-formatted result, and displays it. It contains significant authorization and data-exposure considerations because arbitrary SQL, database selection, role selection, and tokens are accepted, but no clear malicious payload, credential theft, suspicious destination, persistence, or sabotage behavior is present. The supplied fragment is also incomplete or nonfunctional due to multiple undefined identifiers. Review the omitted helper implementations and enforce read-only SQL and role restrictions if this tool is intended for general use.

This module is an authenticated admin CLI for modifying and publishing Apollo test configuration. It does not show strong indicators of embedded malware (no obfuscation, no eval/dynamic execution, no stealth/exfiltration to unknown domains), but it is high-impact and security-sensitive: it transmits a session Cookie to perform privileged writes/releases, can persist that cookie to disk, and logs configuration values that may include secrets. The biggest risks are credential handling, potential secret leakage in logs/CI, and reliance on HTTP/port-80 transport details handled in requestHttp.

This fragment is an overlay UI that persistently stores its own position/config and—most importantly—triggers a sensitive cookie-export capability in extension code via chrome.runtime.sendMessage({action:'exportCookies', gatewayUrl, domain}). The UI forwards user-controlled gatewayUrl and domain without validation here, so safety depends on the receiving extension handler (allowlisting, user consent, and destination enforcement). This module itself shows no obfuscation or direct exfiltration, but it strongly enables a privacy-sensitive workflow and should be reviewed together with the action handler implementation.

This fragment is operational documentation for Apollo configuration management and does not itself contain executable malware. It does describe handling a privileged authentication cookie and mutation/release operations, and it uses plain HTTP internal endpoints, creating meaningful credential and configuration-exposure risks. The production safeguards and explicit confirmation step are positive controls. Review the referenced scripts, protect the cookie file, and require HTTPS or equivalent authenticated transport before use.

This module has strong session/credential-theft characteristics: it reads browser cookies for an arbitrary user-specified/derived domain and exports raw cookie values via clipboard, downloadable files, and a network POST to a configurable gateway endpoint. While it could be intended for legitimate local cookie synchronization/dev tooling, the implemented capability matches common cookie-harvesting/exfiltration patterns and should be treated as high security risk unless the gateway and usage context are strictly controlled.

No direct malware payload is present in the supplied documentation, but it specifies a high-sensitivity credential collection and persistence workflow. Sending complete browser cookies to an AI and storing reusable authentication tokens locally creates a substantial account-takeover and data-disclosure risk. Review the referenced scripts before use, avoid sharing cookies with external systems, minimize token scope, protect the storage file, and use an approved secret-management mechanism.

The manifest contains no direct evidence of malware or data exfiltration because it has no executable logic. It does establish a broad and sensitive capability set: all-site access, cookie access, debugger control, page injection, and externally connectable localhost communication. These capabilities warrant review of the implementation and strict validation of external messages, but the manifest alone is insufficient to classify the extension as malicious.

No direct malware can be established from the supplied documentation. The stated cookie-export behavior is intentionally sensitive and presents a security risk if cookie scope, gateway authentication, local endpoint access, or output-file permissions are weak. Source-code review of the extension manifest and implementation is required to determine whether it performs unauthorized collection or exfiltration.