Back to skill

Security audit

Acp Remote

Security checks for vulnerabilities and agentic risk

Overview

The skill is openly about remote agent command execution, but its install and automation instructions create review-worthy risk by running unverified external code and encouraging blanket approval for remote commands.

Install only if you trust the upstream imclaw and acpx publishers and are comfortable reviewing or replacing the installation steps with pinned, verified artifacts. Avoid --approve-all except in a tightly controlled sandbox, prefer wss:// with authentication, restrict --allowed-tools and --cwd, and do not store sensitive tokens in shell startup files.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (4)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:26
Finding

Execution of an Unpinned Remote Installation Script

Content
View full analysis
Remediation
View remediation
/scripts/install.sh" ``` 4. Publish and verify a cryptographic SHA-256 checksum or a trusted release signature before execution. 5. Execute the installer with the lowest-privileged account possible. 6. Prefer a package manager or a manually verified binary installation process over a remotely mutable installation script. ]]>

T08 · Insecure Dependencies

Error
Location
SKILL.md:47
Finding

Unverified Release Binary Installation and Unsafe Temporary Path

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
SKILL.md:20
Finding

Unpinned Global npm Dependency Installation

Content
View full analysis
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Error
Location
SKILL.md:151
Finding

Blanket Approval Enables Excessive Remote Agent Privileges

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Rogue AgentSelf-Modification, Session Persistence
Findings (11)

Chaining Abuse

High
Category
Tool Misuse
Confidence
99% confidence
Finding

The | bash construct chains network retrieval directly into shell execution, eliminating any opportunity for inspection and magnifying the impact of upstream compromise. In a skill intended for broad reuse, this is a high-risk pattern because it normalizes arbitrary code execution during setup.

Content

Scanner excerpt · SKILL.md (reported line 26)May include surrounding context.

Install imclaw-cli

if ! command -v imclaw-cli &> /dev/null; then curl -fsSL https://raw.githubusercontent.com/smallnest/imclaw/main/scripts/install.sh | bash fi

text

YARA rule 'backdoor_persistence': Backdoor persistence with malicious payloads (shell commands, SSH key injection, hidden root users) [malware]

High
Category
YARA Match
Confidence
75% confidence
Finding

YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).

Content

Scanner excerpt · SKILL.md (reported line 61)May include surrounding context.

on LATEST=$(curl -s https://api.github.com/repos/smallnest/imclaw/releases/latest | grep '"tag_name"' | sed -E 's/."([^"]+)"./\1/')

3. Download and extract

curl -sL "https://github.com/smallnest/imclaw/releases/download/${LATEST}/imclaw_${OS}_${ARCH}.tar.gz" | tar xz -C /tmp

4. Install to PATH

mkdir -p ~/bin mv /tmp/imclaw-cli ~/bin/ chmod +x ~/bin/imclaw-cli

5. Ensure ~/bin is in PATH

echo 'export PATH="$HOME/bin:$PATH"' >> ~/.bashrc # or ~/.zshrc

text

### Verify installation

```bash
imclaw-cli --help
acpx --version

Configuration

Set the remote server URL and optional auth token:

bash
# Default server (can be omitted if using default)
export IMCLAW_SERVER="ws://your-server:8080/ws"

# Auth token (if server requires authentication)
export IMCLAW_TOKEN="your-secret-token"

Or pass directly via command line.

Server Configuration

IMClaw server uses command line flags (no config file needed):

bash
# Start server with default settings
imclaw

# Custom

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
99% confidence
Finding

The installation flow pipes a remotely fetched script directly into bash, which executes network-delivered code without verification or inspection. If the upstream repository, transport, or account is compromised, users could run arbitrary code immediately on their systems.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 50)May include surrounding context.

md
[ "$ARCH" = "aarch64" ] && ARCH="arm64"

# 2. Get latest release version
LATEST=$(curl -s https://api.github.com/repos/smallnest/imclaw/releases/latest | grep '"tag_name"' | sed -E 's/.*"([^"]+)".*/\1/')

# 3. Download and extract
curl -sL "https://github.com/smallnest/imclaw/releases/download/${LATEST}/imclaw_${OS}_${ARCH}.tar.gz" | tar xz -C /tmp

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% 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 · SKILL.md (reported line 56)May include surrounding context.

md
curl -sL "https://github.com/smallnest/imclaw/releases/download/${LATEST}/imclaw_${OS}_${ARCH}.tar.gz" | tar xz -C /tmp

# 4. Install to PATH
mkdir -p ~/bin
mv /tmp/imclaw-cli ~/bin/
chmod +x ~/bin/imclaw-cli

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

--approve-all enables autonomous approval of all permission requests for a remote agent session. In a skill centered on remote ACP command execution, this can let the connected agent run sensitive actions without interactive confirmation, increasing blast radius if the remote server or prompt is unsafe.

Content

Scanner excerpt · SKILL.md (reported line 145)May include surrounding context.

md
| `--session` | Session ID to continue |
| `--agent` | Agent type: claude, codex, etc. (default: claude) |
| `--format` | Output format: text, json, quiet (default: text) |
| `--approve-all` | Auto-approve all permission requests |
| `--approve-reads` | Auto-approve reads, prompt for writes (default) |
| `--deny-all` | Deny all permission requests |
| `--allowed-tools` | Comma-separated tool names (default: Bash,Read,Write) |

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The documentation explicitly recommends --approve-all for a tool that connects to a remote ACP server and can invoke tools such as Bash, Read, and Write. Removing approval gates for a remote agent materially increases the chance of unintended command execution, file access, or modification without user review, especially because the skill’s purpose is remote command execution.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 146)May include surrounding context.

md
| `--agent` | Agent type: claude, codex, etc. (default: claude) |
| `--format` | Output format: text, json, quiet (default: text) |
| `--approve-all` | Auto-approve all permission requests |
| `--approve-reads` | Auto-approve reads, prompt for writes (default) |
| `--deny-all` | Deny all permission requests |
| `--allowed-tools` | Comma-separated tool names (default: Bash,Read,Write) |
| `--cwd` | Working directory |

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% 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 · SKILL.md (reported line 175)May include surrounding context.

Use with SSH tunnel

bash
# Create SSH tunnel first
ssh -L 8080:localhost:8080 user@remote-server -N &

# Connect via localhost

External Script Fetching

Low
Category
Supply Chain
Confidence
98% confidence
Finding

The skill instructs users to fetch an external install script from GitHub and execute it. External script fetching without integrity verification creates a direct supply-chain execution path if the source repository, branch, or delivery path is tampered with.

Content

Scanner excerpt · SKILL.md (reported line 26)May include surrounding context.

Install imclaw-cli

if ! command -v imclaw-cli &> /dev/null; then curl -fsSL https://raw.githubusercontent.com/smallnest/imclaw/main/scripts/install.sh | bash fi

text

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

Appending to ~/.bashrc or ~/.zshrc changes the user's shell environment persistently, affecting future sessions. The skill documentation does not explicitly warn that this modifies user configuration files or suggest reviewing the change before applying it.

Content

No source excerpt is available for this finding.