Secure OSS Dependencies
Beyond CVE Scanning
Open source code makes up more than 90% of modern software projects, with many apps spamming 10,000+ dependencies. This makes it easy for attackers to use open source as a vector for attacks where open source packages registries are frequently the target of malware. Traditional vulnerability scanners cannot detect active supply chain attacks. Socket's free GitHub app safeguards your open source code from both vulnerable and malicious dependencies.

github.com/domren/jingwei
v1.6.1
Live on go
Blocked by Socket
This fragment implements remote installation and management of Java in-memory shells through an encrypted JSP shell, including HTTP-listening agent shells that can provide remote access to a target JVM. The install path transmits a shell key and endpoint configuration, and URL construction can target a supplied host and port; no hardcoded domains, credentials, or local data theft are present. This is high-risk dual-use functionality, and its appropriateness depends on authorized use; severity is high if used to deploy an unauthorized backdoor.
@infituit/opencode-acp
2.0.0-alpha.12
by GitHub Actions
Live on npm
Blocked by Socket
The code implements an OpenCode context-management plugin, but its conditional auto-update path can recursively delete the project directory containing node_modules when auto-update is enabled and a newer eligible package version is detected. The affected path is derived as the parent of node_modules; the update check contacts registry.npmjs.org and the shown code does not reinstall the package. This is a high-severity destructive behavior, although its activation depends on configuration and version conditions.
github.com/rivitna/malware
v0.0.0-20261119194249-5ef0e4728447
Live on go
Blocked by Socket
This module is a high-impact web-content tampering script: it removes CGI help JS assets, backs up and overwrites the public index.html with decompressed embedded data, and then hardens the result with immutable attributes to resist remediation. Even though the snippet shows no explicit command-and-control or credential theft, its behavior aligns strongly with website defacement/persistence or an attack deployment step. Treat as a supply-chain/content compromise indicator requiring immediate containment and integrity verification of the affected web root and assets.
github.com/domren/jingwei
v1.6.6
Live on go
Blocked by Socket
High severity: this module generates and controls encrypted PHP, JSP, and ASPX webshells that execute remote commands, access and modify server files, disclose runtime metadata, rotate keys, and self-delete. Randomized identifiers, decoy comments, carrier variation, timing jitter, and 404 camouflage are concealment indicators. PHP persistence uses an auto_prepend_file variant; temporary sidecar files have build-derived names under the system temporary directory. The code contains no fixed domain or IP; requests target a caller-supplied shell URL. Its capabilities are consistent with authorized administration tooling only when deliberately deployed and controlled, but unauthorized deployment poses an extreme security threat.
github.com/rivitna/malware
v0.0.0-20261119194249-5ef0e4728447
Live on go
Blocked by Socket
This code is highly characteristic of Hive ransomware operational tooling: it parses ransomware-specific encrypted/key filename metadata, derives XOR key material from encrypted/original file pairs (or key files), decrypts by XORing file blocks and overwriting content on disk, and can optionally delete encrypted inputs. The absence of filesystem path restrictions in this fragment further increases the risk of harmful file targeting by callers. Overall, it should be treated as malicious/high-risk code in a supply chain context.
github.com/righettod/toolbox-pentest-web
v0.0.0-20261115002015-cc1504fcf994
Live on go
Blocked by Socket
This code creates poc.zip containing a path-traversal entry targeting /var/www/html/webshell.php and a PHP webshell that executes the c query parameter via system(). The code does not itself extract or run the payload, but the archive can cause command execution if processed by a vulnerable extractor and the PHP file is served. This is an exploit artifact with high security risk; no network access or credential collection is present.
bla-agents
3.5.7
by GitHub Actions
Live on npm
Blocked by Socket
The code modifies pcap data by appending synthetic traffic and does not itself transmit data or access credentials. Its `bankrupt` mode intentionally adds ten near-maximum-size DNS answer packets and tracks estimated tshark JSON output, creating a substantial resource-exhaustion risk for downstream processing. The unusual `/bla/tarpit/exhaust` HTTP path and high-volume headers reinforce the potential for downstream disruption. Severity is high when untrusted captures are processed downstream with `bankrupt` intensity; no external domains, URLs, or IPs are contacted.
github.com/rivitna/malware
v0.0.0-20261119194249-5ef0e4728447
Live on go
Blocked by Socket
High likelihood of malicious ransomware behavior. The code renames a target file to include an encoded marker and a caller-provided “ransom_ext”, then encrypts the file contents in-place using XOR with keytable-derived bytes. It also supports exporting/encrypting the key material with RSA public keys (key escrow), and includes an Erase() method to clear keys from memory. No benign use context or safeguards are present in this fragment.
github.com/infinity6-ai/gox/brwjsz
v0.0.76
Live on go
Blocked by Socket
High-severity browser control code: commands received from `/api/brwjs/events` are evaluated as JavaScript in the page context, and results or errors are sent to `/api/brwjs/reply`. This permits whoever can control or inject commands through that endpoint to act with the page's browser privileges and potentially access or transmit page data. The fragment shows no independent authentication, so its safety depends on protections outside this code; malicious intent cannot be established from this fragment alone.
signed-manifest-guard
0.1.7
Live on pypi
Blocked by Socket
Malicious code in signed-manifest-guard (PyPI) Source: kam193 (f48a108083017729348a2db82bf0102f98051b8e564cee056e353deeaedb2fb2) This package is prepared to be embedded with malicious code, following previous similar packages that eventually were delivering a RAT. --- Category: MALICIOUS - The campaign has clearly malicious intent, like infostealers. Campaign: 2026-11-testing1345 Reasons (based on the campaign): - The package overrides the install command in setup.py to execute malicious code during installation. - Downloads and executes a remote malicious script. - rat
github.com/Mohamed-Rafraf/Challenges_2021_Public
v0.0.0-20210926102117-ade8f3725b75
Live on go
Blocked by Socket
No overt stealth malware (e.g., backdoor/RCE/credential exfiltration) is explicitly present in this single module. However, it contains multiple high-severity application security flaws: an admin login bypass via a token in the GET query string, an internal URL-forwarding mechanism to another localhost service that creates an SSRF/internal-fetch gadget surface, and suspicious Markup-based handling of database-derived usernames that could enable stored XSS depending on template rendering behavior. Additionally, hardcoded SECRET_KEY and disabled HttpOnly for session cookies materially increase exploit impact. This should be treated as high security risk and not used as-is in production.
github.com/Mohamed-Rafraf/Challenges_2021_Public
v0.0.0-20210926102117-ade8f3725b75
Live on go
Blocked by Socket
The program reads stdin with unsafe gets into a 16-byte stack buffer, enabling stack memory corruption. It contains a conditional system("/bin/sh") call gated by a magic value in an otherwise zero-initialized, unused variable, which may be reachable through the overflow depending on build and stack layout. This is a high-severity security risk and a potential shell backdoor; no external network activity or credential collection is present.
github.com/farbodbj/subfinder
v0.0.0-20261019132149-487f2c044abe
Live on go
Blocked by Socket
The core behavior is an explicit “local file -> external messaging” pipeline: it reads the first 10 lines of a locally specified file and transmits them to Telegram using a bot token/target chat configured via .env. This is strongly consistent with an information-disclosure/exfiltration use case if the file content is sensitive. Additionally, sourcing `.env` with `source` introduces a separate high-risk condition where a tampered `.env` could execute arbitrary shell commands on the host. No obvious obfuscation or system-damaging behavior is present in this fragment, but the data egress pattern itself is the dominant security concern.
kage-baileys
25.0.9
by xrljosedv1
Live on npm
Blocked by Socket
High severity availability and integrity concern: the unconditional, hoisted __importStar override causes module initialization to throw, preventing normal use and resembling sabotage. If execution proceeds in a modified or partial context, uploadFile() sends media to catbox[.]moe, scdn[.]pdi[.]moe, qu[.]ax, uguu[.]se, or tmpfiles[.]org, and vid2jpg() contacts ezgif[.]com; these are explicit third-party transfers that may expose user media. The embedded duplicate contains an exec-based ffmpeg command-injection risk, and URL-derived requests require validation. No credential theft or environment-variable collection is evident.
github.com/domren/jingwei
v1.6.3
Live on go
Blocked by Socket
High severity: this code executes caller-supplied commands through configured webshell or in-memory-shell endpoints and returns their output. Its system-information probes collect extensive target details, including environment variables that may contain credentials, and expose results through the API and database. The fragment shows no independent exfiltration destination or obfuscation, and includes URL blocking and timeout controls; however, authorization controls are not visible here, so exposure to unauthorized callers would enable remote command execution and sensitive-data disclosure.
saql4l
1.0.5
Live on pypi
Blocked by Socket
This code is a covertly presented desktop AI assistant that continuously reads clipboard text and can capture screen content, transmitting both to https://meraaaiisseehhapi-erserver[.]vercel[.]app/api; the hidden UI, background relaunch, and global hooks increase the risk of unconsented disclosure. It also has a self-destruct shortcut that deletes uv directories under LOCALAPPDATA and APPDATA, selected matching files under TEMP, and the uv cache before exiting. Severity is high for privacy and system integrity because sensitive clipboard and screen data can leave the machine and local files can be deleted; the fragment does not show credential-specific targeting or an exploit payload.
info@sfer.dev
2.0.10
by SFer
Live on firefox
Blocked by Socket
Malicious extension detected by Annex
augmentedvalueacceleration.ava-supernova
0.121.0
by AugmentedValueAcceleration
Live on vscode
Blocked by Socket
This fragment is predominantly PDF.js parsing and image/decompression code, but its Brotli decoder initialization contains unusually large Caesar-shifted and otherwise encoded string data that appears capable of altering decoder tables. No clear credential theft, exfiltration, reverse shell, or hardcoded suspicious domain is visible. The anomalous data is a credible sabotage or concealed-payload indicator and warrants investigation of the complete file and comparison against the upstream PDF.js build; severity is high but intent and impact are not established from this excerpt alone.
github.com/ValidatorStack/solana-vigil
v0.2.0
Live on go
Blocked by Socket
This code provides a root-only mechanism to install and silently run a caller-supplied shell script from /etc/vigil/bootstrap.sh, then create an installation marker. Detached execution and the persistent installation path are suspicious and could enable unauthorized actions, but the payload is absent, so its effects cannot be established. Severity is high pending inspection of the supplied script; no domains, URLs, IPs, or credential targets are visible.
github.com/IcewhaleTech/CasaOS
v0.1.4
Live on go
Blocked by Socket
This module contains a high-suspicion hardcoded login backdoor-like credential check (admin/admin) that directly mints and returns a JWT token. It also forwards attacker-controlled profile fields and file uploads into user_service methods with no validation in this layer, and returns a token from config.UserInfo.Token in the UserInfo response. There are also potential information-disclosure behaviors via returning raw error messages. Overall, the strongest malicious indicator is the explicit admin/admin backdoor.
instantscrape
0.1.0
Live on pypi
Blocked by Socket
Malicious code in instantscrape (PyPI) Source: kam193 (080884e1b90e6717f2e13d434608203eb24db65a5071803022d3d12b6a7032d5) During installation or import, package executes an embedded malicious executable acting as an infostealer. --- Category: MALICIOUS - The campaign has clearly malicious intent, like infostealers. Campaign: 2026-11-instantscrape Reasons (based on the campaign): - infostealer - The package overrides the install command in setup.py to execute malicious code during installation. - files-exfiltration - obfuscation
github.com/domren/jingwei
v1.6.6
Live on go
Blocked by Socket
This code enables password-based remote shell execution and explicitly describes using weak credentials for internal access, information gathering, and maintaining access. It disables SSH host-key verification, permitting server impersonation and interception. No hardcoded credentials, external domains, or local data theft are present in this fragment; the principal concern is its explicit offensive use framing and unrestricted remote command execution. Severity is high if used without authorization.
instantscrape
0.1.8
Live on pypi
Blocked by Socket
This Windows-only code conditionally places a bundled resource at `~/python.exe` and requests that it run with administrator privileges, using reversed strings for the marker and destination names. This is a significant suspicious behavior and could enable privileged execution of an unknown payload; the payload and `move_resource` implementation are unavailable, so malicious intent cannot be confirmed from this fragment alone. No domains, URLs, IP addresses, or credential targets are present.
github.com/domren/jingwei
v1.4.2
Live on go
Blocked by Socket
This code is a management client for installing and operating Java in-memory shells through an existing webshell, including a standalone HTTP agent and self-removal. That functionality can establish covert remote access and is a high-risk backdoor deployment capability, although this fragment alone does not establish unauthorized use or contain a hardcoded target, credential, or exfiltration destination. No specific domains, IPs, or file paths are hardcoded; the agent URL branch uses HTTP. Severity is high if used without authorization.
github.com/domren/jingwei
v1.2.0
Live on go
Blocked by Socket
This code generates and operates PHP/JSP webshells that accept encrypted HTTP commands and execute them on the server using `system()` or `/bin/sh -c`; its comments explicitly describe hiding command traffic from detection. The default credential is `jingwei`, and the payload also embeds any supplied password; the client transmits that password and command to the supplied URL. This is high-severity malicious-capable functionality, although deployment or use is not shown in this fragment.
github.com/domren/jingwei
v1.6.1
Live on go
Blocked by Socket
This fragment implements remote installation and management of Java in-memory shells through an encrypted JSP shell, including HTTP-listening agent shells that can provide remote access to a target JVM. The install path transmits a shell key and endpoint configuration, and URL construction can target a supplied host and port; no hardcoded domains, credentials, or local data theft are present. This is high-risk dual-use functionality, and its appropriateness depends on authorized use; severity is high if used to deploy an unauthorized backdoor.
@infituit/opencode-acp
2.0.0-alpha.12
by GitHub Actions
Live on npm
Blocked by Socket
The code implements an OpenCode context-management plugin, but its conditional auto-update path can recursively delete the project directory containing node_modules when auto-update is enabled and a newer eligible package version is detected. The affected path is derived as the parent of node_modules; the update check contacts registry.npmjs.org and the shown code does not reinstall the package. This is a high-severity destructive behavior, although its activation depends on configuration and version conditions.
github.com/rivitna/malware
v0.0.0-20261119194249-5ef0e4728447
Live on go
Blocked by Socket
This module is a high-impact web-content tampering script: it removes CGI help JS assets, backs up and overwrites the public index.html with decompressed embedded data, and then hardens the result with immutable attributes to resist remediation. Even though the snippet shows no explicit command-and-control or credential theft, its behavior aligns strongly with website defacement/persistence or an attack deployment step. Treat as a supply-chain/content compromise indicator requiring immediate containment and integrity verification of the affected web root and assets.
github.com/domren/jingwei
v1.6.6
Live on go
Blocked by Socket
High severity: this module generates and controls encrypted PHP, JSP, and ASPX webshells that execute remote commands, access and modify server files, disclose runtime metadata, rotate keys, and self-delete. Randomized identifiers, decoy comments, carrier variation, timing jitter, and 404 camouflage are concealment indicators. PHP persistence uses an auto_prepend_file variant; temporary sidecar files have build-derived names under the system temporary directory. The code contains no fixed domain or IP; requests target a caller-supplied shell URL. Its capabilities are consistent with authorized administration tooling only when deliberately deployed and controlled, but unauthorized deployment poses an extreme security threat.
github.com/rivitna/malware
v0.0.0-20261119194249-5ef0e4728447
Live on go
Blocked by Socket
This code is highly characteristic of Hive ransomware operational tooling: it parses ransomware-specific encrypted/key filename metadata, derives XOR key material from encrypted/original file pairs (or key files), decrypts by XORing file blocks and overwriting content on disk, and can optionally delete encrypted inputs. The absence of filesystem path restrictions in this fragment further increases the risk of harmful file targeting by callers. Overall, it should be treated as malicious/high-risk code in a supply chain context.
github.com/righettod/toolbox-pentest-web
v0.0.0-20261115002015-cc1504fcf994
Live on go
Blocked by Socket
This code creates poc.zip containing a path-traversal entry targeting /var/www/html/webshell.php and a PHP webshell that executes the c query parameter via system(). The code does not itself extract or run the payload, but the archive can cause command execution if processed by a vulnerable extractor and the PHP file is served. This is an exploit artifact with high security risk; no network access or credential collection is present.
bla-agents
3.5.7
by GitHub Actions
Live on npm
Blocked by Socket
The code modifies pcap data by appending synthetic traffic and does not itself transmit data or access credentials. Its `bankrupt` mode intentionally adds ten near-maximum-size DNS answer packets and tracks estimated tshark JSON output, creating a substantial resource-exhaustion risk for downstream processing. The unusual `/bla/tarpit/exhaust` HTTP path and high-volume headers reinforce the potential for downstream disruption. Severity is high when untrusted captures are processed downstream with `bankrupt` intensity; no external domains, URLs, or IPs are contacted.
github.com/rivitna/malware
v0.0.0-20261119194249-5ef0e4728447
Live on go
Blocked by Socket
High likelihood of malicious ransomware behavior. The code renames a target file to include an encoded marker and a caller-provided “ransom_ext”, then encrypts the file contents in-place using XOR with keytable-derived bytes. It also supports exporting/encrypting the key material with RSA public keys (key escrow), and includes an Erase() method to clear keys from memory. No benign use context or safeguards are present in this fragment.
github.com/infinity6-ai/gox/brwjsz
v0.0.76
Live on go
Blocked by Socket
High-severity browser control code: commands received from `/api/brwjs/events` are evaluated as JavaScript in the page context, and results or errors are sent to `/api/brwjs/reply`. This permits whoever can control or inject commands through that endpoint to act with the page's browser privileges and potentially access or transmit page data. The fragment shows no independent authentication, so its safety depends on protections outside this code; malicious intent cannot be established from this fragment alone.
signed-manifest-guard
0.1.7
Live on pypi
Blocked by Socket
Malicious code in signed-manifest-guard (PyPI) Source: kam193 (f48a108083017729348a2db82bf0102f98051b8e564cee056e353deeaedb2fb2) This package is prepared to be embedded with malicious code, following previous similar packages that eventually were delivering a RAT. --- Category: MALICIOUS - The campaign has clearly malicious intent, like infostealers. Campaign: 2026-11-testing1345 Reasons (based on the campaign): - The package overrides the install command in setup.py to execute malicious code during installation. - Downloads and executes a remote malicious script. - rat
github.com/Mohamed-Rafraf/Challenges_2021_Public
v0.0.0-20210926102117-ade8f3725b75
Live on go
Blocked by Socket
No overt stealth malware (e.g., backdoor/RCE/credential exfiltration) is explicitly present in this single module. However, it contains multiple high-severity application security flaws: an admin login bypass via a token in the GET query string, an internal URL-forwarding mechanism to another localhost service that creates an SSRF/internal-fetch gadget surface, and suspicious Markup-based handling of database-derived usernames that could enable stored XSS depending on template rendering behavior. Additionally, hardcoded SECRET_KEY and disabled HttpOnly for session cookies materially increase exploit impact. This should be treated as high security risk and not used as-is in production.
github.com/Mohamed-Rafraf/Challenges_2021_Public
v0.0.0-20210926102117-ade8f3725b75
Live on go
Blocked by Socket
The program reads stdin with unsafe gets into a 16-byte stack buffer, enabling stack memory corruption. It contains a conditional system("/bin/sh") call gated by a magic value in an otherwise zero-initialized, unused variable, which may be reachable through the overflow depending on build and stack layout. This is a high-severity security risk and a potential shell backdoor; no external network activity or credential collection is present.
github.com/farbodbj/subfinder
v0.0.0-20261019132149-487f2c044abe
Live on go
Blocked by Socket
The core behavior is an explicit “local file -> external messaging” pipeline: it reads the first 10 lines of a locally specified file and transmits them to Telegram using a bot token/target chat configured via .env. This is strongly consistent with an information-disclosure/exfiltration use case if the file content is sensitive. Additionally, sourcing `.env` with `source` introduces a separate high-risk condition where a tampered `.env` could execute arbitrary shell commands on the host. No obvious obfuscation or system-damaging behavior is present in this fragment, but the data egress pattern itself is the dominant security concern.
kage-baileys
25.0.9
by xrljosedv1
Live on npm
Blocked by Socket
High severity availability and integrity concern: the unconditional, hoisted __importStar override causes module initialization to throw, preventing normal use and resembling sabotage. If execution proceeds in a modified or partial context, uploadFile() sends media to catbox[.]moe, scdn[.]pdi[.]moe, qu[.]ax, uguu[.]se, or tmpfiles[.]org, and vid2jpg() contacts ezgif[.]com; these are explicit third-party transfers that may expose user media. The embedded duplicate contains an exec-based ffmpeg command-injection risk, and URL-derived requests require validation. No credential theft or environment-variable collection is evident.
github.com/domren/jingwei
v1.6.3
Live on go
Blocked by Socket
High severity: this code executes caller-supplied commands through configured webshell or in-memory-shell endpoints and returns their output. Its system-information probes collect extensive target details, including environment variables that may contain credentials, and expose results through the API and database. The fragment shows no independent exfiltration destination or obfuscation, and includes URL blocking and timeout controls; however, authorization controls are not visible here, so exposure to unauthorized callers would enable remote command execution and sensitive-data disclosure.
saql4l
1.0.5
Live on pypi
Blocked by Socket
This code is a covertly presented desktop AI assistant that continuously reads clipboard text and can capture screen content, transmitting both to https://meraaaiisseehhapi-erserver[.]vercel[.]app/api; the hidden UI, background relaunch, and global hooks increase the risk of unconsented disclosure. It also has a self-destruct shortcut that deletes uv directories under LOCALAPPDATA and APPDATA, selected matching files under TEMP, and the uv cache before exiting. Severity is high for privacy and system integrity because sensitive clipboard and screen data can leave the machine and local files can be deleted; the fragment does not show credential-specific targeting or an exploit payload.
info@sfer.dev
2.0.10
by SFer
Live on firefox
Blocked by Socket
Malicious extension detected by Annex
augmentedvalueacceleration.ava-supernova
0.121.0
by AugmentedValueAcceleration
Live on vscode
Blocked by Socket
This fragment is predominantly PDF.js parsing and image/decompression code, but its Brotli decoder initialization contains unusually large Caesar-shifted and otherwise encoded string data that appears capable of altering decoder tables. No clear credential theft, exfiltration, reverse shell, or hardcoded suspicious domain is visible. The anomalous data is a credible sabotage or concealed-payload indicator and warrants investigation of the complete file and comparison against the upstream PDF.js build; severity is high but intent and impact are not established from this excerpt alone.
github.com/ValidatorStack/solana-vigil
v0.2.0
Live on go
Blocked by Socket
This code provides a root-only mechanism to install and silently run a caller-supplied shell script from /etc/vigil/bootstrap.sh, then create an installation marker. Detached execution and the persistent installation path are suspicious and could enable unauthorized actions, but the payload is absent, so its effects cannot be established. Severity is high pending inspection of the supplied script; no domains, URLs, IPs, or credential targets are visible.
github.com/IcewhaleTech/CasaOS
v0.1.4
Live on go
Blocked by Socket
This module contains a high-suspicion hardcoded login backdoor-like credential check (admin/admin) that directly mints and returns a JWT token. It also forwards attacker-controlled profile fields and file uploads into user_service methods with no validation in this layer, and returns a token from config.UserInfo.Token in the UserInfo response. There are also potential information-disclosure behaviors via returning raw error messages. Overall, the strongest malicious indicator is the explicit admin/admin backdoor.
instantscrape
0.1.0
Live on pypi
Blocked by Socket
Malicious code in instantscrape (PyPI) Source: kam193 (080884e1b90e6717f2e13d434608203eb24db65a5071803022d3d12b6a7032d5) During installation or import, package executes an embedded malicious executable acting as an infostealer. --- Category: MALICIOUS - The campaign has clearly malicious intent, like infostealers. Campaign: 2026-11-instantscrape Reasons (based on the campaign): - infostealer - The package overrides the install command in setup.py to execute malicious code during installation. - files-exfiltration - obfuscation
github.com/domren/jingwei
v1.6.6
Live on go
Blocked by Socket
This code enables password-based remote shell execution and explicitly describes using weak credentials for internal access, information gathering, and maintaining access. It disables SSH host-key verification, permitting server impersonation and interception. No hardcoded credentials, external domains, or local data theft are present in this fragment; the principal concern is its explicit offensive use framing and unrestricted remote command execution. Severity is high if used without authorization.
instantscrape
0.1.8
Live on pypi
Blocked by Socket
This Windows-only code conditionally places a bundled resource at `~/python.exe` and requests that it run with administrator privileges, using reversed strings for the marker and destination names. This is a significant suspicious behavior and could enable privileged execution of an unknown payload; the payload and `move_resource` implementation are unavailable, so malicious intent cannot be confirmed from this fragment alone. No domains, URLs, IP addresses, or credential targets are present.
github.com/domren/jingwei
v1.4.2
Live on go
Blocked by Socket
This code is a management client for installing and operating Java in-memory shells through an existing webshell, including a standalone HTTP agent and self-removal. That functionality can establish covert remote access and is a high-risk backdoor deployment capability, although this fragment alone does not establish unauthorized use or contain a hardcoded target, credential, or exfiltration destination. No specific domains, IPs, or file paths are hardcoded; the agent URL branch uses HTTP. Severity is high if used without authorization.
github.com/domren/jingwei
v1.2.0
Live on go
Blocked by Socket
This code generates and operates PHP/JSP webshells that accept encrypted HTTP commands and execute them on the server using `system()` or `/bin/sh -c`; its comments explicitly describe hiding command traffic from detection. The default credential is `jingwei`, and the payload also embeds any supplied password; the client transmits that password and command to the supplied URL. This is high-severity malicious-capable functionality, although deployment or use is not shown in this fragment.
Proactively search and detect dependencies across repositories in your organization, with actionable insights for your projects and SBOMs

Block emerging malware threats, including intentionally maintainer-added updates, along with packages that differ in name by only a few characters..
Get alerted when a dependency update introduces new risky API usage - filesystem, network, child_process, eval().
Detect obfuscated, minified, or hidden code.
Socket detects the sudden inclusion of a new maintainer, updates with telemetry or protestware added, dependencies pulled in from a remote git URL, and much more.
We help security teams work more efficiently
Get actionable alerts for the supply chain risks that matter. Socket highlights risky dependencies directly within the developer workflow.