Back to skill

Security audit

OpenClaw Boot Camp

Security checks for vulnerabilities and agentic risk

Overview

This skill generates an OpenClaw CLI reference file, and its local writes and optional docs lookup are aligned with that purpose.

Install only if you want a persistent OpenClaw CLI reference in your OpenClaw workspace. Use local-only mode for no network access; use enrich mode only if you are comfortable fetching documentation from docs.openclaw.ai and incorporating it into a file your agent may read later.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (8)

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The declared purpose says the skill generates a CLI reference from the installed version, but the observed behavior expands beyond that into network access, writing completion data, hard-coded operational content, and interactive wizard flows. This mismatch is dangerous because reviewers and agents may grant trust or permissions based on the narrow description while the actual implementation performs broader actions with greater privacy and integrity impact.

Content

No source excerpt is available for this finding.

External Script Fetching

High
Category
Supply Chain
Confidence
94% confidence
Finding

The script fetches remote content from docs.openclaw.ai and processes it into a trusted local reference document. Although it does not execute the fetched content as code, it still imports untrusted network data into an artifact intended to guide an agent's future behavior, creating a supply-chain and prompt-injection risk if the documentation site or network path is compromised.

Content

Scanner excerpt · bootcamp.sh (reported line 361)May include surrounding context.

sh
fetch_doc_flags() {
  local url="$1"
  # --connect-timeout + --max-time prevent the wizard hanging on a flaky doc server
  curl -sL --connect-timeout 5 --max-time 15 "$url" 2>/dev/null | python3 -c "
import sys, re
html = sys.stdin.read()
text = re.sub(r'<script[^>]*>.*?</script>', '', html, flags=re.DOTALL)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
92% confidence
Finding

The skill advertises and instructs shell execution (bash bootcamp.sh) but does not declare any explicit tool scope or permissions boundary. This is dangerous because an agent or user cannot reliably tell, from the manifest alone, that the skill requires command execution and may perform filesystem or other side effects.

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 · README.md (reported line 50)May include surrounding context.

md
echo ""
      echo "Options:"
      echo "  --enrich    Run Phase 2: agent-assisted enrichment from docs"
      echo "  --yes, -y   Non-interactive mode (overwrite existing, skip prompts)"
      echo "  --output=   Custom output directory (default: ~/.openclaw/workspace/notes)"
      echo "  --help, -h  Show this help"
      exit 0

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 · bootcamp.sh (reported line 37)May include surrounding context.

sh
echo ""
      echo "Options:"
      echo "  --enrich    Run Phase 2: agent-assisted enrichment from docs"
      echo "  --yes, -y   Non-interactive mode (overwrite existing, skip prompts)"
      echo "  --output=   Custom output directory (default: ~/.openclaw/workspace/notes)"
      echo "  --help, -h  Show this help"
      exit 0

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 · bootcamp.sh (reported line 37)May include surrounding context.

sh
echo ""
      echo "Options:"
      echo "  --enrich    Run Phase 2: agent-assisted enrichment from docs"
      echo "  --yes, -y   Non-interactive mode (overwrite existing, skip prompts)"
      echo "  --output=   Custom output directory (default: ~/.openclaw/workspace/notes)"
      echo "  --help, -h  Show this help"
      exit 0

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The UI explicitly claims 'No APIs or tokens are used during this process,' but the script can later fetch live content from docs.openclaw.ai when enrichment is enabled. This is misleading security signaling: users may consent under false assumptions and expose network metadata or process untrusted remote content without realizing it.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill writes output into the user's home directory, but the description does not clearly warn about that side effect up front. While lower severity, undisclosed writes can still surprise users, overwrite expected state, or create persistence artifacts that an agent may rely on without the user's informed consent.

Content

No source excerpt is available for this finding.