Back to skill

Security audit

quick-tunnel

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious with medium confidence because its tunnel and chat workflows raise concerns about public exposure, message access, and service persistence.

Treat this as a review decision, not a claim that the author is malicious. Before installing, consider whether you are comfortable making a local service publicly reachable, whether chat messages could be accessed or altered by others with the URL, and whether tunnels might restart across sessions. Scanner coverage was partial because some referenced paths were missing, so the evidence does not establish that every referenced component was reviewed.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (10)

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/architecture.md (reported line 24)May include surrounding context.

text

- **executor.sh** `iniciar|matar NOME`: sobe serviço local (se porta 127.0.0.1
  não responde) com `setsid nohup`, grava `run/N.svc`; sobe cloudflared
  `tunnel --url <redacted-url> --no-autoupdate`, grava `run/N.pid`,
  log em `logs/N.log`.
- **monitor.sh** `validar|estado NOME`: extrai URL do banner do log (regex,

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · scripts/install.sh (reported line 15)May include surrounding context.

sh
echo "[2/4] Binário cloudflared"
if [ -x $B ]; then echo "  ok: $B"; else
  echo "  faltando — baixando em background (cloudflared-linux-amd64)"
  setsid nohup bash -c 'curl -sL -C - -o /home/z/my-project/tunnel/bin/cloudflared.part https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64 && chmod +x /home/z/my-project/tunnel/bin/cloudflared.part && mv /home/z/my-project/tunnel/bin/cloudflared.part /home/z/my-project/tunnel/bin/cloudflared' >/home/z/my-project/tunnel/logs/download.log 2>&1 &
  echo "  download em bg; acompanhe: tail tunnel/logs/download.log"
fi

The start workflow exposes a local service through Cloudflare without warning about public access or third-party traffic handling.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The workflow launches the local service and cloudflared, and the service-setup instructions later tell the agent to test it with tunnelctl start PORTA. Although the document mentions exposed ports, it does not warn that the tunnel makes local content publicly reachable through a third party; port allowlisting and HTTP/content validation do not establish access control.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
83% confidence
Finding

The localhost curl request is not itself external exfiltration, but the documented chat API is published through the tunnel and specifies only a user-chosen name, with no authentication for reading history or posting messages. Under this documented interface, anyone obtaining the public URL could read conversations or impersonate a participant, and attacker messages could reach the agent through its polling workflow.

Content

Scanner excerpt · references/chat-service.md (reported line 19)May include surrounding context.

md
# 2. Usuário abre a URL no celular → a página pede um nome → envia mensagens.

# 3. Agente lê as mensagens novas (micro-comando):
curl -s "http://127.0.0.1:8000/messages?since=N"
#    N = último id visto; retorna JSON com mensagens após N.

# 4. Agente responde:

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/platform-limits.md (reported line 12)May include surrounding context.

md
("broken session 403").
- **Comandos bash >~300 bytes** ou **>~15s em foreground** podem quebrar a
  sessão no mesmo turno.
- **VM reciclada entre turnos**: processos em background (`setsid nohup ... &`)
  MORREM; filesystem `/home/z/my-project` PERSISTE (binário, skill, estado).
- **Chamada bash longa EM PRIMEIRO PLANO com timeout 600000 funcionou**
  (pipeline de ~8 min passou) — foreground longo é mais confiável que

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/platform-limits.md (reported line 12)May include surrounding context.

md
("broken session 403").
- **Comandos bash >~300 bytes** ou **>~15s em foreground** podem quebrar a
  sessão no mesmo turno.
- **VM reciclada entre turnos**: processos em background (`setsid nohup ... &`)
  MORREM; filesystem `/home/z/my-project` PERSISTE (binário, skill, estado).
- **Chamada bash longa EM PRIMEIRO PLANO com timeout 600000 funcionou**
  (pipeline de ~8 min passou) — foreground longo é mais confiável que

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/platform-limits.md (reported line 36)May include surrounding context.

md
## Padrões que FUNCIONARAM

- Script persistido + foreground com timeout 600000 (8 min OK).
- `setsid nohup` para daemons DENTRO de um turno + watchdog que re-cria no
  turno seguinte (auto-cura via autostart/revive).
- Estado em arquivos simples (`desired.txt`, pidfiles, `.url`) em vez de
  memória volátil de processo.

Tunnel recovery instructions omit a warning that restarting tunnels exposes local services through public URLs.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The new-session checklist directs the agent to revive tunnels from persistent desired state before obtaining current confirmation, while the documented recovery machinery restarts entries marked on. This can republish a service after a session or VM boundary when its contents or the user's expectations have changed; reporting the URL afterward is not advance notice of renewed public exposure.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
65% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · scripts/install.sh (reported line 15)May include surrounding context.

sh
echo "[2/4] Binário cloudflared"
if [ -x $B ]; then echo "  ok: $B"; else
  echo "  faltando — baixando em background (cloudflared-linux-amd64)"
  setsid nohup bash -c 'curl -sL -C - -o /home/z/my-project/tunnel/bin/cloudflared.part https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64 && chmod +x /home/z/my-project/tunnel/bin/cloudflared.part && mv /home/z/my-project/tunnel/bin/cloudflared.part /home/z/my-project/tunnel/bin/cloudflared' >/home/z/my-project/tunnel/logs/download.log 2>&1 &
  echo "  download em bg; acompanhe: tail tunnel/logs/download.log"
fi