Skills/npm/ssh-mcp

ssh-mcp@2.26.0

CAUTION

Publisher claims: "MCP server exposing policy-gated, audited SSH access for Linux and Windows hosts via the Model Context Protocol." Observed in sandbox: no network, touches nothing outside the project, writes no memory.

Registry npm (MCP server)Publisher tufantunc · publishing for 567 daysInstalls 53,841Last detonated 38 h ago
What would fix thisWatchDispute verdict

Coverage

Server started and answered tools/list.
Tested
MCP initialize and tools/list · every tool description, for hidden instructions · every file read, file written and host contacted by what ran
Not tested
what the tools do when called with real credentials

Before you give it your token

It will ask for
SSH_AUTH_SOCKSSH_MCP_PASSWORDSSH_MCP_SUDO_PASSWORD or --bearertokenConfiguration, not credentials: MAX_REMOTE_PATH_CHARS, RESERVED_BINDING_KEYS, SSH_MCP_DISABLE_MAIN, SSH_MCP_KEY, SSH_MCP_PASSPHRASE, SSH_MCP_TEST, SystemRoot, USERNAME
Where it talks to
no vendor host named in the code
What the honeypot saw
no credential files read · no network
What we could not test
what the tools do when called with real credentials

Connect it with a scoped credential, under the guard. It asks for SSH_AUTH_SOCK, SSH_MCP_PASSWORD, SSH_MCP_SUDO_PASSWORD and talks to its vendor. Give the credential to the guard, not the server: the value never enters the server's environment and only travels to the host you bind.

$vet-skills run --net declared -- npx -y ssh-mcp

Never paste a credential into a chat, a ticket or a report, including ours. How the guard holds it.

Is [email protected] safe to install?

KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios) on 2026-11-12 and scored it 63/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Destructive file system command detected; Destructive command against a filesystem root or device; Destructive file system command detected. Summary: no network, touches nothing outside the project, writes no memory.

Behavior trace

Honeypot run · 37 s · 2 scenarios
t+CategoryObserved action
0.0sprocessno install scripts in bundle
16sprocessdependency install exit 0
36sfilesystemread ~/.config/ssh-mcp/config.toml outside project
36sclaimed15 tools: list-connections, list-sessions, open-session, close-session, read-session-output, read-command, run-command, privileged-command, signal-process, sftp-upload, sftp-download, sftp-list, sftp-upload-file, sftp-download-file, reload-config
36sprocessscript backend runs setup only
36sprocessno install scripts to re-run

Static findings

HIGH
Destructive file system command detectedbuild/policy/classifier.js:628 · Destructive file system command detected// spelling, and `rm -fr /` or `mkfs -t ext4 /dev/sda1` classified `safe` —Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
CRITICAL
Destructive command against a filesystem root or devicebuild/policy/classifier.js:632 · Destructive command against a filesystem root or device// regex refused any substring, so `echo 'rm -rf /'` and a grep quotingFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Destructive file system command detectedbuild/policy/classifier.js:991 · Destructive file system command detected* /"` prints a string, and normalising it produced `echo hello; rm -rf /`, which matchedFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/policy/awk.js:332 · Privileged sudo command — agent may execute with elevated privileges// gawk's indirect call (`f="system"; @f("sudo id")`) and its `@load` /Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/policy/awk.js:97 · Privileged sudo command — agent may execute with elevated privileges* `BEGIN{not=2; x = not / system("sudo id") / 3}` is a *division* that real awkFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Destructive file system command detectedbuild/guard/approval-grants.js:43 · Destructive file system command detected* `rm -rf /tmp/build` must not also approve `rm -rf /tmp/build-prod`.Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Destructive file system command detectedbuild/policy/classifier.js:632 · Destructive file system command detected// regex refused any substring, so `echo 'rm -rf /'` and a grep quotingFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Access to sensitive file pathsbuild/guard/sanitizer.js:28 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files* `read-command "-c sftp:download /etc/shadow"` resolved to binaryFix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
HIGH
Privileged sudo commandbuild/guard/sanitizer.js:13 · Privileged sudo command — agent may execute with elevated privileges* `read-command "sftp:list /tmp sudo id"` and the classifier answeredFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/policy/awk.js:355 · Privileged sudo command — agent may execute with elevated privileges// Measured: `print "x" | "id" "; sudo sh"` came out `safe`.Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/policy/awk.js:376 · Privileged sudo command — agent may execute with elevated privileges// getline` runs `sudo id`, and reading only the last literal classifiedFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
CRITICAL
Pipe-to-shell patternbuild/policy/classifier.js:1743 · Pipe-to-shell pattern — downloads and immediately executes remote code* empty part between them, and this deliberately still matches: `curl x || sh`Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST06
HIGH
Privileged sudo commandbuild/policy/classifier.js:872 · Privileged sudo command — agent may execute with elevated privileges* anything that reads arguments by position: `awk -F '' 'BEGIN{system("sudoFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Destructive file system command detectedbuild/policy/classifier.js:543 · Destructive file system command detected// classifier never looked at — `env sudo rm -f /etc/passwd` classifiedFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
CRITICAL
Reads credential paths from the home directorybuild/policy/classifier.js:1918 · Reads credential paths from the home directory* `break`-on-close reading as a bypass: `> ""~/.ssh/authorized_keys` fell through).Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST03
HIGH
Privileged sudo commandbuild/guard/sanitizer.js:122 · Privileged sudo command — agent may execute with elevated privileges* by taking the *higher* class — so `/tmp/x; sudo id` classifies `privileged`Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/policy/classifier.js:125 · Privileged sudo command — agent may execute with elevated privileges* `find` can carry `operandsAreData: true` while `find … -exec sudo id +` isFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/policy/classifier.js:566 · Privileged sudo command — agent may execute with elevated privileges* `sftp:list /tmp sh -c 'sudo id'` fell from `privileged` to `read-only`,Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/policy/classifier.js:543 · Privileged sudo command — agent may execute with elevated privileges// classifier never looked at — `env sudo rm -f /etc/passwd` classifiedFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Destructive file system command detectedbuild/policy/classifier.js:625 · Destructive file system command detected// The four operation rules that used to live here — `rm -rf /`, `mkfs…`,Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
CRITICAL
Destructive command against a filesystem root or devicebuild/policy/classifier.js:1845 · Destructive command against a filesystem root or device* Device formatters and wipers: any `mkfs` spelling — `mkfs.ext4 /dev/sda1`Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Access to sensitive file pathsbuild/policy/classifier.js:514 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files// `uniq /etc/passwd /root/.ssh/authorized_keys`; GNU's own usage line isFix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
File permission modification commandpackage.json:13 · File permission modification command — agent changes file ownership/permissions"build": "tsc && shx chmod +x build/*.js",Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
Process kill commandbuild/ssh/connection.js:435 · Process kill command — agent can terminate processes// count by as long as the kill ladder takes.Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/tools/command-tools.js:71 · Privileged sudo command — agent may execute with elevated privilegeswrap: (cleanCmd) => `sudo -p "" -S sh -c ${shellSingleQuote(`exec 0</dev/null; ${cleanCmd}`)}`,Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/tools/command-tools.js:64 · Privileged sudo command — agent may execute with elevated privileges// written for it stays in the pipe, and sudo exec's this `sh -c` withFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/tools/command-tools.js:62 · Privileged sudo command — agent may execute with elevated privileges// `sudo -S` reads stdin only when it actually needs a password. UnderFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
File permission modification commandbuild/policy/classifier.js:608 · File permission modification command — agent changes file ownership/permissions* literal head matches, so a command built from `dd curl wget chown -R x `Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
HIGH
Access to sensitive file pathsbuild/tools/forbidden-dirs.js:23 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files* - a root shaped like `~/.ssh` turns the same two tools into "read the privateFix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
File permission modification commandbuild/policy/classifier.js:626 · File permission modification command — agent changes file ownership/permissions// `iptables -F`, `chmod -R 777 /` — are segment checks in FORBIDDEN_RULESFix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
HIGH
Privileged sudo commandbuild/tools/pipeline.js:300 · Privileged sudo command — agent may execute with elevated privileges// The sudo password is the one secret that can plausibly end up in thisFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Privileged sudo commandbuild/policy/engine.js:438 · Privileged sudo command — agent may execute with elevated privileges* sudo and goes looking for a bigger role that does not exist (#91).Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
File permission modification commandbuild/config/loader.js:64 · File permission modification command — agent changes file ownership/permissions`Required: 0700. Run: chmod 700 ${dir}\n` +Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
HIGH
Access to sensitive file pathsbuild/tools/forbidden-dirs.js:29 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files{ path: join(homedir(), '.ssh'), reason: 'the SSH client directory (~/.ssh)' },Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
HIGH
Privileged sudo commandbuild/tools/command-tools.js:67 · Privileged sudo command — agent may execute with elevated privileges// before the command runs means sudo still gets its password from theFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Process kill commandbuild/ssh/connection-registry.js:243 · Process kill command — agent can terminate processes// for its sessions' kill ladders, and serialising multiplied that wait by the number ofFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
File permission modification commandbuild/policy/classifier.js:1798 · File permission modification command — agent changes file ownership/permissions* out of reach (documented residual). Measured in review: `chmod -R 777Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
File permission modification commandbuild/config/loader.js:67 · File permission modification command — agent changes file ownership/permissions// with no way to chmod a path baked into the image.Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
HIGH
Access to sensitive file pathsbuild/policy/classifier.js:582 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files* profile nothing new: `cat /etc/shadow` and `ls /root` are already `read-only`,Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
HIGH
Privileged sudo commandbuild/tools/command-tools.js:50 · Privileged sudo command — agent may execute with elevated privileges// Evaluate the bare command too: the sudo wrapper would otherwiseFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
HIGH
Access to sensitive file pathsbuild/tools/local-path.js:135 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files// of the transfer that wrote it, and a root shaped like ~/.ssh turns the sameFix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
HIGH
Access to sensitive file pathsbuild/policy/classifier.js:117 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files* an accepted — reader past `safe` (`sort -o /tmp/x /etc/passwd` isFix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
File permission modification commandbuild/ssh/sftp.js:401 · File permission modification command — agent changes file ownership/permissions* bit are dropped rather than carried across. See the chmod below.Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
HIGH
Privileged sudo commandbuild/policy/classifier.js:949 · Privileged sudo command — agent may execute with elevated privileges// sudo id` then never splits on the `;` and the elevation disappears, which measured asFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Process kill commandbuild/tools/command-tools.js:83 · Process kill command — agent can terminate processesreturn runAudited(`kill -${signal} ${pid}`, { toolName: 'signal-process', failureClass: 'destructive', profile, extra, synthetic: true }, async (rt) => {Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Process kill commandbuild/guard/sanitizer.js:35 · Process kill command — agent can terminate processes* `kill -TERM <pid>`, which is outside this namespace and cannot be reserved —Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
File permission modification commandbuild/config/loader.js:24 · File permission modification command — agent changes file ownership/permissions* `chmod 600` it prescribed could not lift that rejection — measured on Windows 11,Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
MCP server using network transport (SSE/HTTP)build/index.js:58 · MCP server using network transport (SSE/HTTP) — data sent over network// An McpServer binds to a single transport, so HTTP needs one per session.Fix: Namespace tool names and keep tool descriptions literal; no instructions to the agent inside them.
AST04
HIGH
Privileged sudo commandbuild/tools/descriptions.js:10 · Privileged sudo command — agent may execute with elevated privileges…-upload and run it by path.', 'privileged-command': 'Execute a command with sudo elevation. Goes through the approval gate; approvalPolicy on the profile decides…Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Unicode escape sequence detectedbuild/guard/sanitizer.js:99 · Unicode escape sequence detected — may contain obfuscated payloadsexport const PATH_CONTROL_CHARS = /[\u0000-\u001f\u007f-\u009f\u061c\u200b\u200e\u200f\u2028-\u202e\u2060-\u2064\u2066-\u2069\ufeff]/;Fix: Put hosts, commands and strings in plain source. Decoding at runtime looks identical to malware and costs points on its own.
AST07
MEDIUM
Process kill commandbuild/index.js:119 · Process kill command — agent can terminate processes// runtime's SIGKILL. Teardown is concurrent now, so the honest ceiling is one killFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Process kill commandbuild/ssh/session-manager.js:149 · Process kill command — agent can terminate processes// Guarded and in a `finally`, both because closing now awaits a kill ladder ratherFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Process kill commandbuild/index.js:110 · Process kill command — agent can terminate processes// behind a teardown that now waits for kill ladders. Flushing first cannot lose a recordFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
Process kill commandbuild/index.js:93 · Process kill command — agent can terminate processes* One kill ladder (3s) plus slack. Sessions and connections close concurrently, so thisFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
File permission modification commandpackage.json:14 · File permission modification command — agent changes file ownership/permissions"build:coverage": "tsc --sourceMap && shx chmod +x build/*.js",Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
HIGH
Access to sensitive file pathsbuild/policy/classifier.js:543 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files// classifier never looked at — `env sudo rm -f /etc/passwd` classifiedFix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
Process kill commandbuild/ssh/connection-registry.js:189 · Process kill command — agent can terminate processes* `toClose`, unclosed: closing awaits session kill ladders, and the caller does it afterFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
File permission modification commandbuild/config/windows-acl.js:346 · File permission modification command — agent changes file ownership/permissions// the same defect as the `chmod 600` this whole change replaced.Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
File permission modification commandbuild/config/loader.js:59 · File permission modification command — agent changes file ownership/permissions`Required: 0600. Run: chmod 600 ${filePath}`);Fix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
File permission modification commandbuild/policy/classifier.js:1785 · File permission modification command — agent changes file ownership/permissions* takes both `-r` and `-R`, while chmod and chown take only `-R` — lowercaseFix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
HIGH
Access to sensitive file pathsbuild/policy/classifier.js:552 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files// (`curl -d @/etc/passwd`), and write remote files (`curl -o`, `wget -O`) —Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
Process kill commandbuild/tools/results.js:9 · Process kill command — agent can terminate processes* Render a CommandResult for the client. A non-zero exit (or a kill signal)Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05

Cost to the agent

Instructions loaded
≈62 tok
1 file · 58 in bundle
Tool list per prompt
≈2,508 tok
15 tools · 167 per tool
Handshake
742 ms
launch to tools/list, stdio
Undocumented tools
0
description under 20 chars · avg 228

Tokens estimated at four characters each. The tool list is re-read by the agent on every turn, so it is the number to compare between servers that do the same job: ranked across the index.

Known vulnerabilities

No advisories in OSV for ssh-mcp@2.26.0 as of 38 h ago. This is the package itself, not its dependencies.

What would make this Verified

17 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%
  1. Reads credential files≈ +27 on the totalThe sandbox saw reads under ~/.ssh, ~/.aws, ~/.npmrc, ~/.config or similar. Read only files inside the project, and only the ones the task needs. Seen: ~/.config/ssh-mcp/config.toml.
  2. Destructive command against a filesystem root or device≈ +9 on the totalBuild commands from fixed arguments; do not interpolate untrusted strings into a shell.
  3. Pipe-to-shell pattern≈ +9 on the totalBuild commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
  4. Reads credential paths from the home directory≈ +9 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
  5. Unclaimed publisher≈ +8 on the totalClaim tufantunc: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
  6. Touches files outside the project≈ +7 on the totalKeep reads and writes inside the working directory, or declare the exact paths in `filesystem:` and explain why in the README.
Publish a new version to npm, then queue it here. Not your skill? Claim the handle to get verdict changes by email.Claim tufantuncPropose a patch to the maintainer