Secure your dependencies. Ship with confidence.

This module fragment includes a high-severity arbitrary code execution mechanism. It dynamically generates a Web Worker and then evaluates attacker-controlled serialized function code (eval on both the main thread for validation and inside the worker for actual execution) and invokes the resulting functions with attacker-controlled arguments. In any supply-chain scenario where job payloads can be influenced, this functions as a backdoor-like capability and should be treated as a critical security risk.

This code fragment is strongly consistent with malicious telemetry/exfiltration: it collects hostname and user information from the local environment and unconditionally sends them to a hardcoded external endpoint over HTTPS on module load. It contains no legitimate business logic, no consent/config gating, and no sanitization/redaction.

This code fragment is strongly consistent with malicious telemetry/exfiltration: it collects hostname and user information from the local environment and unconditionally sends them to a hardcoded external endpoint over HTTPS on module load. It contains no legitimate business logic, no consent/config gating, and no sanitization/redaction.

`lotusbail` is a malicious npm package that masquerades as a WhatsApp Web API library by forking legitimate Baileys-based code and preserving working messaging functionality. In addition to normal API behavior, it inserts a wrapper around the WhatsApp WebSocket client so that all traffic passing through the library is duplicated for collection. Reported data theft includes WhatsApp authentication tokens and session keys, full message content (sent/received and historical), contact lists (including phone numbers), and transferred media/files. The package also attempts to establish persistent unauthorized access by hijacking the WhatsApp device-linking (“pairing”) workflow using a hardcoded pairing code, effectively linking an attacker-controlled device to the victim’s account; removing the npm dependency does not automatically remove the linked device. To hinder detection, the exfiltration endpoint is hidden behind multiple obfuscation layers, collected data is encrypted (including a custom RSA implementation), and the code includes anti-debugging traps designed to disrupt analysis.

This module embeds an HTTP POST side effect that spawns the external `gh auth login --web` command in detached mode with ignored stdio and suppressed visibility. That is atypical for standard Next.js route runtime code and creates a high-risk pathway for unauthorized process execution / credential-adjacent behavior if the route is reachable without strong authorization and auditing. Additionally, it can disclose error details by returning `e.message` on spawn failure. Other parts of the fragment appear to be normal Next.js server scaffolding without obvious additional malicious primitives.

This module contains explicit client-side exfil/notification behavior: it hardcodes Feishu/Lark bot webhook URLs (test/prod) and sends dynamically constructed messages containing caller-provided title/content via fetch POST. It also includes clipboard-write capabilities (navigator.clipboard + execCommand fallback) and logs configuration used for environment/webhook selection. Even if intended for legitimate internal notifications, embedding bot webhook endpoints and transmitting dynamic content externally is a high-risk supply-chain characteristic and should be reviewed for legitimacy, user consent, and secret handling/redaction.

High security risk. This module injects a browser-side script into proxied HTML that listens for postMessage events and performs arbitrary JavaScript evaluation using new Function(text) with attacker-controlled inputs. It then posts results/errors back to the parent window using postMessage with wildcard origin, providing a direct RCE + data exfiltration primitive in the client context. Coupled with host-side command execution helpers, privileged /etc/hosts modification, and disabled TLS verification for outbound proxying, the overall supply-chain/abuse potential is substantial and warrants immediate review/removal or strict hardening (origin allowlisting, message authentication, and eliminating dynamic evaluation).

This code fragment is strongly consistent with malicious telemetry/exfiltration: it collects hostname and user information from the local environment and unconditionally sends them to a hardcoded external endpoint over HTTPS on module load. It contains no legitimate business logic, no consent/config gating, and no sanitization/redaction.

The provided fragment includes a high-risk design: a Web Worker job system that reconstructs and executes caller-provided functions using eval() in both the main thread and the worker. It also dynamically generates worker script code via a Blob URL. If an attacker can influence job.doFunction or job.args[].funcString (directly or via untrusted data paths), this becomes arbitrary code execution within the application’s JS context (Worker). Other parts of the fragment are primarily parsing/streaming with some DoS-related bounds, but they do not offset the fundamental eval/Worker execution capability.

High-risk behavior is present: the module provides an arbitrary JavaScript execution engine (`processJS` -> `runJS`) and additionally evaluates user-provided snippet code via `eval(...)` through a Proxy (`snippets`). This strongly elevates the likelihood of supply-chain sabotage or malicious runtime behavior. The same bundle also includes a server-side file read helper (`fs.readFileSync`) in `code.embed`, which could enable local file disclosure if its arguments are attacker-influenced. Overall, the fragment is not merely a utility library; it contains explicit code-execution capabilities consistent with a backdoor/weaponizable feature.

This code is highly consistent with an offensive exploitation framework: it automates payload generation and execution against targets, acquires authenticated sessions by filling login forms and extracting cookies/tokens from the browser context, then uses that session to perform further exploitation. It also records payloads and response previews and publishes “confirmed” findings via telemetry/memoryBank. While safety/kill-switch controls exist, the overall behavior is not benign and presents a high risk of misuse and sensitive data handling.

The code is largely standard for an auth/project API, but it contains a high-risk supply-chain style behavior: _generate_token_setup generates a dynamic `python3 -c` stop-hook command that reads a local transcript file specified by the CLAUDE_SESSION_TRANSCRIPT environment variable and sends the transcript to `${base_url}/api/sessions/analyze` over the network with a Bearer token. This is consistent with data exfiltration/privacy invasion and an execution hook that could be abused for sabotage or unauthorized data collection. If this package is distributed/used broadly, this should be treated as an extremely suspicious/malicious component and reviewed in the associated hook runner/consumer context.

This code is a payload/template generator explicitly designed to craft social-engineering, prompt-injection, data-exfiltration, privilege-escalation, and remote-code-execution payloads. It contains hardcoded malicious endpoints and commands, fabricates authority/confirmation artifacts, and tailors messages to increase success. Treat this module as malicious; investigate repository context and callers, remove and block its use in trusted environments, and audit any systems where it was present for potential abuse.

This code performs targeted credential/token harvesting from Cursor IDE’s local SQLite state database (including accessToken and machineId) and exfiltrates the results by returning them in a network-facing Next.js GET JSON response. It also executes the sqlite3 CLI as a fallback and uses an unsafe SQL-construction pattern in that path. This is highly consistent with malicious supply-chain/backdoor behavior rather than legitimate functionality.

High-risk supply-chain/security indicator: this module implements a WebWorker job executor that uses eval() to reconstruct and execute caller-provided function strings both in the main thread (via eval in TsWorker.convertJobToObj) and inside the generated Worker script (eval of funcString and doFunction). If an attacker can influence job payloads, this becomes arbitrary code execution and a plausible sabotage/backdoor vector. While the surrounding media/ZIP parsing logic appears to include defensive bounds, it is secondary to the explicit dynamic execution mechanism.

This module provides automated periodic desktop screenshot capture and exfiltrates the resulting image content to external systems (Discord webhook endpoints obtained at runtime, or a relay via JSON containing base64 screenshot data). While it includes operational guardrails (interval/queue/size bounds) and basic input validation plus memory-clearing attempts, the core functionality is high-risk for privacy/data exfiltration. No classic malware primitives (eval/Function, shelling out, filesystem writes) are visible here, so intent is not provable from this fragment alone, but the implemented capability is strongly aligned with spyware/exfiltration patterns.

The code fragment contains multiple extremely high-risk behaviors consistent with an attacker backdoor or remote administration tool: it accepts client-supplied commands and executes them (writeToExec), supports arbitrary file reads returning base64 content, and provides a generic outbound network proxy to attacker-controlled URLs (axios). These are direct source-to-sink flows from untrusted inputs to dangerous operations with minimal validation, making the security risk very high if any attacker can access these endpoints/WebSocket messages.

This module combines standard cookie/route utilities with a high-risk cloud synchronization feature. When enabled, it collects sensitive provider API keys and token-bearing credential data, derives a stable host fingerprint (machine-id/hostname), and POSTs that information to a configurable external endpoint. It then consumes the cloud response to update local provider credentials/tokens, creating a two-way secret synchronization channel. While this could be intended for legitimate remote administration, it is also consistent with supply-chain backdoor/exfiltration behavior; review is warranted for destination allowlisting, authentication/authorization of the sync endpoint, and strict minimization of transmitted secrets.

The code presents a severe security risk due to eval on untrusted input within a JSON reviver, enabling arbitrary code execution. To reduce risk, eliminate dynamic execution from the reviver, or replace with strict whitelisting or safe deserialization mechanisms. If dynamic reconstruction is required, restrict to a predefined registry of permitted functions and validate input before invocation.

This module performs direct arbitrary code execution via `exec(cleaned_code, local_scope)` on provided Python source text, with only syntax checking (`ast.parse`) as a “guard.” If an attacker can influence the input string, it is a high-severity supply-chain/runtime code execution risk (RCE) capable of leading to data theft, exfiltration, persistence, or system modification. The use of a shared `local_scope` increases the chance of state contamination across calls.

The DesktopDriver class itself mainly performs UI automation (create controls, wait/restore window state, send hotkeys). However, the same file/assembly includes heavily obfuscated code plus kernel32 interop for OpenProcess/VirtualAlloc/WriteProcessMemory/VirtualProtect and cryptographic/resource deobfuscation logic—patterns strongly associated with payload loaders/injection. This makes the dependency a serious supply-chain security risk. Recommend sandboxing, removing/quarantining, and performing full assembly-level behavior verification (runtime tracing, import analysis, and resource extraction).

This module is strongly oriented toward offensive exploitation planning: it converts vulnerability evidence into structured, multi-stage attack paths with explicit attacker actions (session/token theft, credential exfiltration, SSRF to cloud metadata/IAM credential harvesting, and RCE-to-persistence/backdoor descriptions) and provides a static 'confirmation' heuristic suitable for automation. While actual execution/payload delivery is not present in the shown fragment, the exported/related execution-oriented functions referenced by the module indicate likely end-to-end exploit workflow support elsewhere in the package. Overall, the security risk is high for any production or third-party environment integration.

This module is a RAG FastAPI service with ingestion via subprocess and chat via Ollama + persistent Chroma. The most critical issue is highly suspicious destructive logic in the WebSocket chat handler: it deletes the entire Chroma collection ('zettabrain_docs') and overwrites the ingestion log ('{}'), effectively wiping RAG state. There is no authentication/authorization guarding these actions, and additional signs of snippet corruption/misplaced prompt text further increase the likelihood of tampering. Treat the package as severely compromised/sabotage-capable until the full source (including indentation/trigger conditions) is verified and the destructive behavior is removed/guarded behind authenticated admin controls.

This module is strongly oriented toward offensive exploitation planning: it converts vulnerability evidence into structured, multi-stage attack paths with explicit attacker actions (session/token theft, credential exfiltration, SSRF to cloud metadata/IAM credential harvesting, and RCE-to-persistence/backdoor descriptions) and provides a static 'confirmation' heuristic suitable for automation. While actual execution/payload delivery is not present in the shown fragment, the exported/related execution-oriented functions referenced by the module indicate likely end-to-end exploit workflow support elsewhere in the package. Overall, the security risk is high for any production or third-party environment integration.

This module introduces extremely high-risk capabilities driven by user input: it can execute arbitrary shell commands via execSync when messages begin with '!', and it can read local files based on user-supplied @mention paths (with potential path traversal) and embed the contents into the prompt forwarded to downstream hooks/models. These behaviors create clear pathways for system compromise and sensitive data exfiltration. If this feature is intended, it must be tightly permission-gated and constrained (e.g., remove execSync, add strict command allowlisting, enforce a safe read root/deny traversal, and perform authorization checks before any side-effecting operations).

This module fragment includes a high-severity arbitrary code execution mechanism. It dynamically generates a Web Worker and then evaluates attacker-controlled serialized function code (eval on both the main thread for validation and inside the worker for actual execution) and invokes the resulting functions with attacker-controlled arguments. In any supply-chain scenario where job payloads can be influenced, this functions as a backdoor-like capability and should be treated as a critical security risk.

This code fragment is strongly consistent with malicious telemetry/exfiltration: it collects hostname and user information from the local environment and unconditionally sends them to a hardcoded external endpoint over HTTPS on module load. It contains no legitimate business logic, no consent/config gating, and no sanitization/redaction.

This code fragment is strongly consistent with malicious telemetry/exfiltration: it collects hostname and user information from the local environment and unconditionally sends them to a hardcoded external endpoint over HTTPS on module load. It contains no legitimate business logic, no consent/config gating, and no sanitization/redaction.

`lotusbail` is a malicious npm package that masquerades as a WhatsApp Web API library by forking legitimate Baileys-based code and preserving working messaging functionality. In addition to normal API behavior, it inserts a wrapper around the WhatsApp WebSocket client so that all traffic passing through the library is duplicated for collection. Reported data theft includes WhatsApp authentication tokens and session keys, full message content (sent/received and historical), contact lists (including phone numbers), and transferred media/files. The package also attempts to establish persistent unauthorized access by hijacking the WhatsApp device-linking (“pairing”) workflow using a hardcoded pairing code, effectively linking an attacker-controlled device to the victim’s account; removing the npm dependency does not automatically remove the linked device. To hinder detection, the exfiltration endpoint is hidden behind multiple obfuscation layers, collected data is encrypted (including a custom RSA implementation), and the code includes anti-debugging traps designed to disrupt analysis.

This module embeds an HTTP POST side effect that spawns the external `gh auth login --web` command in detached mode with ignored stdio and suppressed visibility. That is atypical for standard Next.js route runtime code and creates a high-risk pathway for unauthorized process execution / credential-adjacent behavior if the route is reachable without strong authorization and auditing. Additionally, it can disclose error details by returning `e.message` on spawn failure. Other parts of the fragment appear to be normal Next.js server scaffolding without obvious additional malicious primitives.

This module contains explicit client-side exfil/notification behavior: it hardcodes Feishu/Lark bot webhook URLs (test/prod) and sends dynamically constructed messages containing caller-provided title/content via fetch POST. It also includes clipboard-write capabilities (navigator.clipboard + execCommand fallback) and logs configuration used for environment/webhook selection. Even if intended for legitimate internal notifications, embedding bot webhook endpoints and transmitting dynamic content externally is a high-risk supply-chain characteristic and should be reviewed for legitimacy, user consent, and secret handling/redaction.

High security risk. This module injects a browser-side script into proxied HTML that listens for postMessage events and performs arbitrary JavaScript evaluation using new Function(text) with attacker-controlled inputs. It then posts results/errors back to the parent window using postMessage with wildcard origin, providing a direct RCE + data exfiltration primitive in the client context. Coupled with host-side command execution helpers, privileged /etc/hosts modification, and disabled TLS verification for outbound proxying, the overall supply-chain/abuse potential is substantial and warrants immediate review/removal or strict hardening (origin allowlisting, message authentication, and eliminating dynamic evaluation).

This code fragment is strongly consistent with malicious telemetry/exfiltration: it collects hostname and user information from the local environment and unconditionally sends them to a hardcoded external endpoint over HTTPS on module load. It contains no legitimate business logic, no consent/config gating, and no sanitization/redaction.

The provided fragment includes a high-risk design: a Web Worker job system that reconstructs and executes caller-provided functions using eval() in both the main thread and the worker. It also dynamically generates worker script code via a Blob URL. If an attacker can influence job.doFunction or job.args[].funcString (directly or via untrusted data paths), this becomes arbitrary code execution within the application’s JS context (Worker). Other parts of the fragment are primarily parsing/streaming with some DoS-related bounds, but they do not offset the fundamental eval/Worker execution capability.

High-risk behavior is present: the module provides an arbitrary JavaScript execution engine (`processJS` -> `runJS`) and additionally evaluates user-provided snippet code via `eval(...)` through a Proxy (`snippets`). This strongly elevates the likelihood of supply-chain sabotage or malicious runtime behavior. The same bundle also includes a server-side file read helper (`fs.readFileSync`) in `code.embed`, which could enable local file disclosure if its arguments are attacker-influenced. Overall, the fragment is not merely a utility library; it contains explicit code-execution capabilities consistent with a backdoor/weaponizable feature.

This code is highly consistent with an offensive exploitation framework: it automates payload generation and execution against targets, acquires authenticated sessions by filling login forms and extracting cookies/tokens from the browser context, then uses that session to perform further exploitation. It also records payloads and response previews and publishes “confirmed” findings via telemetry/memoryBank. While safety/kill-switch controls exist, the overall behavior is not benign and presents a high risk of misuse and sensitive data handling.

The code is largely standard for an auth/project API, but it contains a high-risk supply-chain style behavior: _generate_token_setup generates a dynamic `python3 -c` stop-hook command that reads a local transcript file specified by the CLAUDE_SESSION_TRANSCRIPT environment variable and sends the transcript to `${base_url}/api/sessions/analyze` over the network with a Bearer token. This is consistent with data exfiltration/privacy invasion and an execution hook that could be abused for sabotage or unauthorized data collection. If this package is distributed/used broadly, this should be treated as an extremely suspicious/malicious component and reviewed in the associated hook runner/consumer context.

This code is a payload/template generator explicitly designed to craft social-engineering, prompt-injection, data-exfiltration, privilege-escalation, and remote-code-execution payloads. It contains hardcoded malicious endpoints and commands, fabricates authority/confirmation artifacts, and tailors messages to increase success. Treat this module as malicious; investigate repository context and callers, remove and block its use in trusted environments, and audit any systems where it was present for potential abuse.

This code performs targeted credential/token harvesting from Cursor IDE’s local SQLite state database (including accessToken and machineId) and exfiltrates the results by returning them in a network-facing Next.js GET JSON response. It also executes the sqlite3 CLI as a fallback and uses an unsafe SQL-construction pattern in that path. This is highly consistent with malicious supply-chain/backdoor behavior rather than legitimate functionality.

High-risk supply-chain/security indicator: this module implements a WebWorker job executor that uses eval() to reconstruct and execute caller-provided function strings both in the main thread (via eval in TsWorker.convertJobToObj) and inside the generated Worker script (eval of funcString and doFunction). If an attacker can influence job payloads, this becomes arbitrary code execution and a plausible sabotage/backdoor vector. While the surrounding media/ZIP parsing logic appears to include defensive bounds, it is secondary to the explicit dynamic execution mechanism.

This module provides automated periodic desktop screenshot capture and exfiltrates the resulting image content to external systems (Discord webhook endpoints obtained at runtime, or a relay via JSON containing base64 screenshot data). While it includes operational guardrails (interval/queue/size bounds) and basic input validation plus memory-clearing attempts, the core functionality is high-risk for privacy/data exfiltration. No classic malware primitives (eval/Function, shelling out, filesystem writes) are visible here, so intent is not provable from this fragment alone, but the implemented capability is strongly aligned with spyware/exfiltration patterns.

The code fragment contains multiple extremely high-risk behaviors consistent with an attacker backdoor or remote administration tool: it accepts client-supplied commands and executes them (writeToExec), supports arbitrary file reads returning base64 content, and provides a generic outbound network proxy to attacker-controlled URLs (axios). These are direct source-to-sink flows from untrusted inputs to dangerous operations with minimal validation, making the security risk very high if any attacker can access these endpoints/WebSocket messages.

This module combines standard cookie/route utilities with a high-risk cloud synchronization feature. When enabled, it collects sensitive provider API keys and token-bearing credential data, derives a stable host fingerprint (machine-id/hostname), and POSTs that information to a configurable external endpoint. It then consumes the cloud response to update local provider credentials/tokens, creating a two-way secret synchronization channel. While this could be intended for legitimate remote administration, it is also consistent with supply-chain backdoor/exfiltration behavior; review is warranted for destination allowlisting, authentication/authorization of the sync endpoint, and strict minimization of transmitted secrets.

The code presents a severe security risk due to eval on untrusted input within a JSON reviver, enabling arbitrary code execution. To reduce risk, eliminate dynamic execution from the reviver, or replace with strict whitelisting or safe deserialization mechanisms. If dynamic reconstruction is required, restrict to a predefined registry of permitted functions and validate input before invocation.

This module performs direct arbitrary code execution via `exec(cleaned_code, local_scope)` on provided Python source text, with only syntax checking (`ast.parse`) as a “guard.” If an attacker can influence the input string, it is a high-severity supply-chain/runtime code execution risk (RCE) capable of leading to data theft, exfiltration, persistence, or system modification. The use of a shared `local_scope` increases the chance of state contamination across calls.

The DesktopDriver class itself mainly performs UI automation (create controls, wait/restore window state, send hotkeys). However, the same file/assembly includes heavily obfuscated code plus kernel32 interop for OpenProcess/VirtualAlloc/WriteProcessMemory/VirtualProtect and cryptographic/resource deobfuscation logic—patterns strongly associated with payload loaders/injection. This makes the dependency a serious supply-chain security risk. Recommend sandboxing, removing/quarantining, and performing full assembly-level behavior verification (runtime tracing, import analysis, and resource extraction).

This module is strongly oriented toward offensive exploitation planning: it converts vulnerability evidence into structured, multi-stage attack paths with explicit attacker actions (session/token theft, credential exfiltration, SSRF to cloud metadata/IAM credential harvesting, and RCE-to-persistence/backdoor descriptions) and provides a static 'confirmation' heuristic suitable for automation. While actual execution/payload delivery is not present in the shown fragment, the exported/related execution-oriented functions referenced by the module indicate likely end-to-end exploit workflow support elsewhere in the package. Overall, the security risk is high for any production or third-party environment integration.

This module is a RAG FastAPI service with ingestion via subprocess and chat via Ollama + persistent Chroma. The most critical issue is highly suspicious destructive logic in the WebSocket chat handler: it deletes the entire Chroma collection ('zettabrain_docs') and overwrites the ingestion log ('{}'), effectively wiping RAG state. There is no authentication/authorization guarding these actions, and additional signs of snippet corruption/misplaced prompt text further increase the likelihood of tampering. Treat the package as severely compromised/sabotage-capable until the full source (including indentation/trigger conditions) is verified and the destructive behavior is removed/guarded behind authenticated admin controls.

This module is strongly oriented toward offensive exploitation planning: it converts vulnerability evidence into structured, multi-stage attack paths with explicit attacker actions (session/token theft, credential exfiltration, SSRF to cloud metadata/IAM credential harvesting, and RCE-to-persistence/backdoor descriptions) and provides a static 'confirmation' heuristic suitable for automation. While actual execution/payload delivery is not present in the shown fragment, the exported/related execution-oriented functions referenced by the module indicate likely end-to-end exploit workflow support elsewhere in the package. Overall, the security risk is high for any production or third-party environment integration.

This module introduces extremely high-risk capabilities driven by user input: it can execute arbitrary shell commands via execSync when messages begin with '!', and it can read local files based on user-supplied @mention paths (with potential path traversal) and embed the contents into the prompt forwarded to downstream hooks/models. These behaviors create clear pathways for system compromise and sensitive data exfiltration. If this feature is intended, it must be tightly permission-gated and constrained (e.g., remove execSync, add strict command allowlisting, enforce a safe read root/deny traversal, and perform authorization checks before any side-effecting operations).