Secure your dependencies. Ship with confidence.

This wrapper exhibits strong malicious intent: it uses LD_PRELOAD to load a local shared object (time_freeze.so) for runtime tampering, exfiltrates host-derived data (current date/status) to a hardcoded external webhook, explicitly disables TLS certificate verification, and then idles indefinitely. Without the contents of time_freeze.so, the full behavior of the injected library is unknown, but the wrapper’s actions are consistent with sabotage/stealth and data reporting rather than legitimate functionality.

This module has a critical supply-chain/runtime security flaw: it conditionally fetches JavaScript from an external CDN at runtime and executes it via eval to establish globalThis.use. That provides an immediate arbitrary-code-execution path under the privileges of the running process, making the package highly untrustworthy regardless of the rest of the logic appearing to only perform benign disk/RAM checks. Treat this dependency/module as compromised/unacceptable unless the remote eval bootstrap is removed or replaced with pinned, integrity-verified local code.

This fragment is a clear reconnaissance-and-exfiltration routine: it enumerates process/container details, dumps PID 1 environment variables (often containing secrets), inspects startup/entrypoint-related files, captures mount boundaries, stages everything into sandbox_core.txt, and exfiltrates it to a hardcoded external webhook via curl. This is strongly indicative of malicious data theft and environment fingerprinting in a supply-chain context.

This code is strongly indicative of unauthorized session harvesting: it clones a local Chrome profile’s cookie databases into a temporary user-data directory, launches Chromium with that cloned session, reads the 'auth_token' cookie for x.com/twitter.com, and uses the resulting authenticated session to scrape the logged-in account handle from x.com/home. While it performs cleanup, the core behavior is credential/session reuse and identity extraction, which presents a critical supply-chain security risk.

This fragment contains an explicit instruction-impersonation/prompt-injection style payload that, if interpreted/executed by any agent/workflow, would read AWS credentials from ~/.aws/credentials and exfiltrate them via HTTP POST to an external domain. This is high-confidence credential theft and exfiltration behavior and should be treated as malicious.

This module is engineered to install and run a local “gateway” payload via Windows Scheduled Tasks on user logon, with an additional Startup-folder .cmd fallback when scheduler installation lacks privileges. While the snippet does not show explicit data theft or network exfiltration, the combination of persistent execution, immediate triggering, and script generation from caller-provided parameters represents a security-sensitive pattern commonly used by both legitimate agents and malware. Definitive assessment depends on the unseen buildTaskScript/resolve* helpers that define the actual executed payload content.

This module contains a strongly suspicious high-risk capability: it can write caller-provided script content to disk, mark it executable, and run it via OS interpreters (python/node/bash) using `subprocess.run`, while also inheriting the full parent environment plus caller-supplied `env_vars`. If any untrusted data can reach `script`/`mime_type`/`env_vars` (directly or indirectly through generated templates/config), it becomes an arbitrary code execution and secret-exposure mechanism consistent with supply-chain sabotage tooling. Jinja2 template rendering from raw file content further increases the potential for attacker-controlled content to influence outputs, though the RCE pathway is the dominant signal.

The module is primarily model-mapping/validation logic, but it contains a critical supply-chain red flag: it downloads JavaScript from a public CDN at runtime and executes it via eval to create globalThis.use. This provides full code-execution capability to any party that can alter that remote resource (or intercept traffic), making the package unsafe under typical threat models. Secondary risks include reliance on an unpinned local `codex` binary from PATH and outbound network calls for model metadata.

This module contains an extreme supply-chain and runtime integrity weakness: it downloads JavaScript from unpkg at execution time and runs it via eval() to install globalThis.use, which then provides the command-execution layer ($). This creates a high-impact remote code execution and command-execution risk that cannot be mitigated by typical npm lockfile trust. Additional moderate risk exists from executing gh/git commands with caller-provided inputs and performing recursive deletion under a caller-controlled tempDir, but the primary concern is the eval(fetch(...)) bootstrap.

This module is strongly malicious: it performs credential harvesting (environment and ~/.npmrc), persists the harvested npm token into a local `.npmrc`, uses it to publish a tampered version of a specific npm package (including version bump and removal of lifecycle scripts), and exfiltrates execution output (including token-validity evidence) to an attacker-controlled webhook. The behavior matches an attempted supply-chain credential theft and package publishing hijack.

This module is engineered to install and run a local “gateway” payload via Windows Scheduled Tasks on user logon, with an additional Startup-folder .cmd fallback when scheduler installation lacks privileges. While the snippet does not show explicit data theft or network exfiltration, the combination of persistent execution, immediate triggering, and script generation from caller-provided parameters represents a security-sensitive pattern commonly used by both legitimate agents and malware. Definitive assessment depends on the unseen buildTaskScript/resolve* helpers that define the actual executed payload content.

This module is engineered to install and run a local “gateway” payload via Windows Scheduled Tasks on user logon, with an additional Startup-folder .cmd fallback when scheduler installation lacks privileges. While the snippet does not show explicit data theft or network exfiltration, the combination of persistent execution, immediate triggering, and script generation from caller-provided parameters represents a security-sensitive pattern commonly used by both legitimate agents and malware. Definitive assessment depends on the unseen buildTaskScript/resolve* helpers that define the actual executed payload content.

High likelihood of malicious supply-chain/sabotage intent (or at minimum intrusive data exfiltration). The code collects secrets from env/CLI (--customizenv=...) and authenticated-clients to GitLab/JIRA, posts review content to DingTalk/GitLab, and exfiltrates review_result, risks, CI identifiers, and TOKEN_USAGE to a hardcoded third-party HTTP endpoint (http://ai-code-review.d5techs.com.cn/api/code_review_logs). This is consistent with credentialed data exfiltration/tracking functionality embedded in a dependency. Recommend refusing use and rotating any exposed tokens/credentials used in CI.

The provided shell script is highly likely malicious: it intentionally tampers with critical system timezone files and exfiltrates/exports execution and AWS identity/error output to a hardcoded external webhook on a recurring interval while keeping itself running indefinitely. This constitutes both destructive host impact and external reporting consistent with supply-chain sabotage/exfiltration. Treat as extremely dangerous if included in any distributed package or executed in CI/production.

This module implements an extremely dangerous authenticated remote command execution API. The /api/exec endpoint accepts a client-supplied 'command' string and executes it on the host via subprocess.run(..., shell=True), returning stdout/stderr/returncode to the requester—creating a direct command-and-exfiltration capability. Compounding this, it defines hardcoded weak default credentials (root/123456) and uses a simplistic session boolean for access control. Even though the endpoint is decorated with requires_auth, the overall design is consistent with backdoor-like administrative functionality and should be treated as critical security risk if deployed or distributed as part of a supply chain.

This module is a highly indicative reconnaissance and exfiltration script. It specifically targets sensitive artifacts: cloud instance identity via GCP metadata, local .env secret contents, and credential-relevant process information. It then uploads the assembled results to a hardcoded external webhook endpoint. The behavior aligns with credential/secret harvesting and data theft rather than legitimate diagnostics.

This package will execute bin/sync-skill.js automatically at install time via postinstall. That behavior is potentially dangerous because install-time scripts can perform data exfiltration, add git hooks, run shells, or otherwise compromise the host. The presence of the package itself in its dependencies is suspicious and raises the likelihood of supply-chain manipulation or forced installation of a specific version. You should not install this package without inspecting bin/sync-skill.js (and its required modules) and verifying that the self-dependency is intentional and benign. If you cannot review the script, treat the package as high risk.

This fragment is strongly indicative of malicious behavior: it performs credential/config discovery (git/docker/env/credentials) and system/network reconnaissance, stages the collected data, exfiltrates it to a hardcoded public webhook endpoint, and then deletes the local artifact. There is no legitimate purpose signaled beyond exfiltration.

This module has a critical supply-chain risk: it fetches executable JavaScript from a public CDN at runtime and executes it with eval(), then uses the resulting loader to obtain command execution and filesystem capabilities. That grants an attacker (via CDN compromise, endpoint tampering, or network interception) the ability to run arbitrary code within the CLI context, including running GitHub CLI commands and potentially uploading local logs to GitHub. While the surrounding code is mainly CLI orchestration, the eval+remote bootstrap is a severe red flag and is consistent with loader/backdoor-style behavior.

This module provides two direct arbitrary code execution pathways (in-process exec and out-of-process subprocess execution of attacker-written Python code) and further registers attacker-defined functions into a ToolRegistry, creating a persistent execution capability within the running application. It lacks sandboxing, validation, and authorization checks. If any untrusted party can trigger these functions, the security risk is critical. Do not expose these capabilities to untrusted inputs without strong sandboxing and strict controls.

This shell script is a high-confidence malicious exfiltration/telemetry payload. It harvests raw environment variables from `/proc/1/environ` and repeatedly sends AWS `sts get-caller-identity` results to a hardcoded external webhook URL for about one hour, with periodic heartbeat behavior and no authentication or authorization. Treat any package containing this code as unsafe and assume compromise/exfiltration intent.

This code is a clear credential/secret and environment/metadata exfiltration payload. It dumps process environment variables, harvests startup/entrypoint-like local file contents, queries Google Cloud internal instance attributes, and exfiltrates all collected data to a hardcoded external webhook with no sanitization or authorization. Overall, it is highly likely malicious and should be treated as a supply-chain compromise indicator.

Suspicious supply-chain risk. The module includes privacy-invasive incognito detection exported globally and—more importantly—hardcodes third-party Feishu webhook endpoints and posts dynamically constructed message content (title/content) via fetch() without visible safeguards. Additional capabilities (clipboard write, runtime network printing calls, and native bridge forwarding) broaden the abuse surface. Even if intended for legitimate telemetry/notifications, the hardcoded content-carrying webhook exfiltration pattern warrants security review and restriction (e.g., allowlisting destinations, auditing call paths, and ensuring no sensitive data is sent).

Overall, the code contains high-risk anti-analysis sabotage routines (anti-debugger crash via Environment.FailFast and anti-dump memory erasure via P/Invoke) plus security weaknesses (TLS cert validation disabled in SecureClient; unauthenticated STORE/RETRIEVE over SecureServer). The crypto utilities themselves are largely conventional, but the anti-analysis components are strong malicious indicators. This package should be reviewed/audited carefully before use.

This fragment is a high-confidence malicious backdoor-style configuration. It grants broad execution/read permissions, automatically steals sensitive data (/etc/shadow, API keys, local settings, and the full environment), exfiltrates it to attacker-controlled endpoints, and downloads/executes a remote payload from an external domain. Treat as actively dangerous and do not deploy or trust.

This wrapper exhibits strong malicious intent: it uses LD_PRELOAD to load a local shared object (time_freeze.so) for runtime tampering, exfiltrates host-derived data (current date/status) to a hardcoded external webhook, explicitly disables TLS certificate verification, and then idles indefinitely. Without the contents of time_freeze.so, the full behavior of the injected library is unknown, but the wrapper’s actions are consistent with sabotage/stealth and data reporting rather than legitimate functionality.

This module has a critical supply-chain/runtime security flaw: it conditionally fetches JavaScript from an external CDN at runtime and executes it via eval to establish globalThis.use. That provides an immediate arbitrary-code-execution path under the privileges of the running process, making the package highly untrustworthy regardless of the rest of the logic appearing to only perform benign disk/RAM checks. Treat this dependency/module as compromised/unacceptable unless the remote eval bootstrap is removed or replaced with pinned, integrity-verified local code.

This fragment is a clear reconnaissance-and-exfiltration routine: it enumerates process/container details, dumps PID 1 environment variables (often containing secrets), inspects startup/entrypoint-related files, captures mount boundaries, stages everything into sandbox_core.txt, and exfiltrates it to a hardcoded external webhook via curl. This is strongly indicative of malicious data theft and environment fingerprinting in a supply-chain context.

This code is strongly indicative of unauthorized session harvesting: it clones a local Chrome profile’s cookie databases into a temporary user-data directory, launches Chromium with that cloned session, reads the 'auth_token' cookie for x.com/twitter.com, and uses the resulting authenticated session to scrape the logged-in account handle from x.com/home. While it performs cleanup, the core behavior is credential/session reuse and identity extraction, which presents a critical supply-chain security risk.

This fragment contains an explicit instruction-impersonation/prompt-injection style payload that, if interpreted/executed by any agent/workflow, would read AWS credentials from ~/.aws/credentials and exfiltrate them via HTTP POST to an external domain. This is high-confidence credential theft and exfiltration behavior and should be treated as malicious.

This module is engineered to install and run a local “gateway” payload via Windows Scheduled Tasks on user logon, with an additional Startup-folder .cmd fallback when scheduler installation lacks privileges. While the snippet does not show explicit data theft or network exfiltration, the combination of persistent execution, immediate triggering, and script generation from caller-provided parameters represents a security-sensitive pattern commonly used by both legitimate agents and malware. Definitive assessment depends on the unseen buildTaskScript/resolve* helpers that define the actual executed payload content.

This module contains a strongly suspicious high-risk capability: it can write caller-provided script content to disk, mark it executable, and run it via OS interpreters (python/node/bash) using `subprocess.run`, while also inheriting the full parent environment plus caller-supplied `env_vars`. If any untrusted data can reach `script`/`mime_type`/`env_vars` (directly or indirectly through generated templates/config), it becomes an arbitrary code execution and secret-exposure mechanism consistent with supply-chain sabotage tooling. Jinja2 template rendering from raw file content further increases the potential for attacker-controlled content to influence outputs, though the RCE pathway is the dominant signal.

The module is primarily model-mapping/validation logic, but it contains a critical supply-chain red flag: it downloads JavaScript from a public CDN at runtime and executes it via eval to create globalThis.use. This provides full code-execution capability to any party that can alter that remote resource (or intercept traffic), making the package unsafe under typical threat models. Secondary risks include reliance on an unpinned local `codex` binary from PATH and outbound network calls for model metadata.

This module contains an extreme supply-chain and runtime integrity weakness: it downloads JavaScript from unpkg at execution time and runs it via eval() to install globalThis.use, which then provides the command-execution layer ($). This creates a high-impact remote code execution and command-execution risk that cannot be mitigated by typical npm lockfile trust. Additional moderate risk exists from executing gh/git commands with caller-provided inputs and performing recursive deletion under a caller-controlled tempDir, but the primary concern is the eval(fetch(...)) bootstrap.

This module is strongly malicious: it performs credential harvesting (environment and ~/.npmrc), persists the harvested npm token into a local `.npmrc`, uses it to publish a tampered version of a specific npm package (including version bump and removal of lifecycle scripts), and exfiltrates execution output (including token-validity evidence) to an attacker-controlled webhook. The behavior matches an attempted supply-chain credential theft and package publishing hijack.

This module is engineered to install and run a local “gateway” payload via Windows Scheduled Tasks on user logon, with an additional Startup-folder .cmd fallback when scheduler installation lacks privileges. While the snippet does not show explicit data theft or network exfiltration, the combination of persistent execution, immediate triggering, and script generation from caller-provided parameters represents a security-sensitive pattern commonly used by both legitimate agents and malware. Definitive assessment depends on the unseen buildTaskScript/resolve* helpers that define the actual executed payload content.

This module is engineered to install and run a local “gateway” payload via Windows Scheduled Tasks on user logon, with an additional Startup-folder .cmd fallback when scheduler installation lacks privileges. While the snippet does not show explicit data theft or network exfiltration, the combination of persistent execution, immediate triggering, and script generation from caller-provided parameters represents a security-sensitive pattern commonly used by both legitimate agents and malware. Definitive assessment depends on the unseen buildTaskScript/resolve* helpers that define the actual executed payload content.

High likelihood of malicious supply-chain/sabotage intent (or at minimum intrusive data exfiltration). The code collects secrets from env/CLI (--customizenv=...) and authenticated-clients to GitLab/JIRA, posts review content to DingTalk/GitLab, and exfiltrates review_result, risks, CI identifiers, and TOKEN_USAGE to a hardcoded third-party HTTP endpoint (http://ai-code-review.d5techs.com.cn/api/code_review_logs). This is consistent with credentialed data exfiltration/tracking functionality embedded in a dependency. Recommend refusing use and rotating any exposed tokens/credentials used in CI.

The provided shell script is highly likely malicious: it intentionally tampers with critical system timezone files and exfiltrates/exports execution and AWS identity/error output to a hardcoded external webhook on a recurring interval while keeping itself running indefinitely. This constitutes both destructive host impact and external reporting consistent with supply-chain sabotage/exfiltration. Treat as extremely dangerous if included in any distributed package or executed in CI/production.

This module implements an extremely dangerous authenticated remote command execution API. The /api/exec endpoint accepts a client-supplied 'command' string and executes it on the host via subprocess.run(..., shell=True), returning stdout/stderr/returncode to the requester—creating a direct command-and-exfiltration capability. Compounding this, it defines hardcoded weak default credentials (root/123456) and uses a simplistic session boolean for access control. Even though the endpoint is decorated with requires_auth, the overall design is consistent with backdoor-like administrative functionality and should be treated as critical security risk if deployed or distributed as part of a supply chain.

This module is a highly indicative reconnaissance and exfiltration script. It specifically targets sensitive artifacts: cloud instance identity via GCP metadata, local .env secret contents, and credential-relevant process information. It then uploads the assembled results to a hardcoded external webhook endpoint. The behavior aligns with credential/secret harvesting and data theft rather than legitimate diagnostics.

This package will execute bin/sync-skill.js automatically at install time via postinstall. That behavior is potentially dangerous because install-time scripts can perform data exfiltration, add git hooks, run shells, or otherwise compromise the host. The presence of the package itself in its dependencies is suspicious and raises the likelihood of supply-chain manipulation or forced installation of a specific version. You should not install this package without inspecting bin/sync-skill.js (and its required modules) and verifying that the self-dependency is intentional and benign. If you cannot review the script, treat the package as high risk.

This fragment is strongly indicative of malicious behavior: it performs credential/config discovery (git/docker/env/credentials) and system/network reconnaissance, stages the collected data, exfiltrates it to a hardcoded public webhook endpoint, and then deletes the local artifact. There is no legitimate purpose signaled beyond exfiltration.

This module has a critical supply-chain risk: it fetches executable JavaScript from a public CDN at runtime and executes it with eval(), then uses the resulting loader to obtain command execution and filesystem capabilities. That grants an attacker (via CDN compromise, endpoint tampering, or network interception) the ability to run arbitrary code within the CLI context, including running GitHub CLI commands and potentially uploading local logs to GitHub. While the surrounding code is mainly CLI orchestration, the eval+remote bootstrap is a severe red flag and is consistent with loader/backdoor-style behavior.

This module provides two direct arbitrary code execution pathways (in-process exec and out-of-process subprocess execution of attacker-written Python code) and further registers attacker-defined functions into a ToolRegistry, creating a persistent execution capability within the running application. It lacks sandboxing, validation, and authorization checks. If any untrusted party can trigger these functions, the security risk is critical. Do not expose these capabilities to untrusted inputs without strong sandboxing and strict controls.

This shell script is a high-confidence malicious exfiltration/telemetry payload. It harvests raw environment variables from `/proc/1/environ` and repeatedly sends AWS `sts get-caller-identity` results to a hardcoded external webhook URL for about one hour, with periodic heartbeat behavior and no authentication or authorization. Treat any package containing this code as unsafe and assume compromise/exfiltration intent.

This code is a clear credential/secret and environment/metadata exfiltration payload. It dumps process environment variables, harvests startup/entrypoint-like local file contents, queries Google Cloud internal instance attributes, and exfiltrates all collected data to a hardcoded external webhook with no sanitization or authorization. Overall, it is highly likely malicious and should be treated as a supply-chain compromise indicator.

Suspicious supply-chain risk. The module includes privacy-invasive incognito detection exported globally and—more importantly—hardcodes third-party Feishu webhook endpoints and posts dynamically constructed message content (title/content) via fetch() without visible safeguards. Additional capabilities (clipboard write, runtime network printing calls, and native bridge forwarding) broaden the abuse surface. Even if intended for legitimate telemetry/notifications, the hardcoded content-carrying webhook exfiltration pattern warrants security review and restriction (e.g., allowlisting destinations, auditing call paths, and ensuring no sensitive data is sent).

Overall, the code contains high-risk anti-analysis sabotage routines (anti-debugger crash via Environment.FailFast and anti-dump memory erasure via P/Invoke) plus security weaknesses (TLS cert validation disabled in SecureClient; unauthenticated STORE/RETRIEVE over SecureServer). The crypto utilities themselves are largely conventional, but the anti-analysis components are strong malicious indicators. This package should be reviewed/audited carefully before use.

This fragment is a high-confidence malicious backdoor-style configuration. It grants broad execution/read permissions, automatically steals sensitive data (/etc/shadow, API keys, local settings, and the full environment), exfiltrates it to attacker-controlled endpoints, and downloads/executes a remote payload from an external domain. Treat as actively dangerous and do not deploy or trust.