Back to skill

Security audit

skill-firewall

Security checks for vulnerabilities and agentic risk

Overview

ClawScan’s verdict is benign with medium confidence: most checks are OK, while the environment proportionality check notes a reported mismatch between the skill’s stated purpose and an installation action; some references were also unresolved, so evidence is incomplete.

The verdict is benign, but it does not mean risk-free. Before installing, consider the reported mismatch: the skill’s instructions include an external install action although its stated purpose is pre-install scanning. The network permission declaration also does not cover a detected network capability, and runtime enforcement is unknown. Two local references were unresolved, so the review evidence is incomplete.

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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Rogue AgentSelf-Modification, Session Persistence
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (19)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied command performs installation, not security analysis. Installing a specific external skill is a materially different operation from the declared pre-installation scanner, and the chunk shows no validation before installation. This mismatch does not establish that the installed skill itself is malicious.

Content

No source excerpt is available for this finding.

Code capability 'network' detected in scripts/skill_firewall.py but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'network' capability that is not covered by the declared 'allowed-tools' preapproval. Review the host runtime's actual enforcement policy; runtime permissions are unknown unless separately verified, and this declaration alone does not establish a permission ceiling.

Content

No source excerpt is available for this finding.

YARA rule 'agent_skill_prompt_injection_hidden_instructions': Prompt injection or hidden instructions embedded in AI agent skill text [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 4)May include surrounding context.

python
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Skill Firewall — static security scanner for AI agent skills (SKILL.md packs).

Scans a skill directory (or any text file) for malicious patterns:
  - Prompt injection / instruction override / role hijack
  - Credential theft & data exfiltration
  - Destructive commands & persistence
  - Obfuscation (base64/hex payloads, zero-width steganography, eval chains)
  - Policy attacks (self-modification, spreading, external instruction loading)

Stdlib only. Deterministic. Safe: never executes any content found in scanned files.

Exit codes: 0=SAFE  1=CAUTION  2=SUSPICIOUS  3=MALICIOUS  4=error
"""

import argparse
import base64

Credential Access

High
Category
Privilege Escalation
Confidence
90% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 122)May include surrounding context.

python
desc="References access to credential stores or secret files",
         patterns=[
             r"~/\.ssh/|/home/[^/\s]+/\.ssh/|/root/\.ssh/|\bid_(rsa|ed25519|ecdsa|dsa)\b",
             r"~/\.aws/credentials|\.aws/credentials|~/\.kube/config|\.kube/config",
             r"~/\.docker/config\.json|\.git-credentials|~/\.netrc|\.npmrc|\.pypirc",
             r"/etc/(shadow|passwd|sudoers)\b|/etc/ssh/sshd_config",
             r"~/\.openclaw/(credentials|identity|agents)/",

Credential Access

High
Category
Privilege Escalation
Confidence
90% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 122)May include surrounding context.

python
desc="References access to credential stores or secret files",
         patterns=[
             r"~/\.ssh/|/home/[^/\s]+/\.ssh/|/root/\.ssh/|\bid_(rsa|ed25519|ecdsa|dsa)\b",
             r"~/\.aws/credentials|\.aws/credentials|~/\.kube/config|\.kube/config",
             r"~/\.docker/config\.json|\.git-credentials|~/\.netrc|\.npmrc|\.pypirc",
             r"/etc/(shadow|passwd|sudoers)\b|/etc/ssh/sshd_config",
             r"~/\.openclaw/(credentials|identity|agents)/",

Credential Access

High
Category
Privilege Escalation
Confidence
80% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 122)May include surrounding context.

python
desc="References access to credential stores or secret files",
         patterns=[
             r"~/\.ssh/|/home/[^/\s]+/\.ssh/|/root/\.ssh/|\bid_(rsa|ed25519|ecdsa|dsa)\b",
             r"~/\.aws/credentials|\.aws/credentials|~/\.kube/config|\.kube/config",
             r"~/\.docker/config\.json|\.git-credentials|~/\.netrc|\.npmrc|\.pypirc",
             r"/etc/(shadow|passwd|sudoers)\b|/etc/ssh/sshd_config",
             r"~/\.openclaw/(credentials|identity|agents)/",

Credential Access

High
Category
Privilege Escalation
Confidence
80% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 122)May include surrounding context.

python
desc="References access to credential stores or secret files",
         patterns=[
             r"~/\.ssh/|/home/[^/\s]+/\.ssh/|/root/\.ssh/|\bid_(rsa|ed25519|ecdsa|dsa)\b",
             r"~/\.aws/credentials|\.aws/credentials|~/\.kube/config|\.kube/config",
             r"~/\.docker/config\.json|\.git-credentials|~/\.netrc|\.npmrc|\.pypirc",
             r"/etc/(shadow|passwd|sudoers)\b|/etc/ssh/sshd_config",
             r"~/\.openclaw/(credentials|identity|agents)/",

Credential Access

High
Category
Privilege Escalation
Confidence
90% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 123)May include surrounding context.

python
patterns=[
             r"~/\.ssh/|/home/[^/\s]+/\.ssh/|/root/\.ssh/|\bid_(rsa|ed25519|ecdsa|dsa)\b",
             r"~/\.aws/credentials|\.aws/credentials|~/\.kube/config|\.kube/config",
             r"~/\.docker/config\.json|\.git-credentials|~/\.netrc|\.npmrc|\.pypirc",
             r"/etc/(shadow|passwd|sudoers)\b|/etc/ssh/sshd_config",
             r"~/\.openclaw/(credentials|identity|agents)/",
         ]),

Credential Access

High
Category
Privilege Escalation
Confidence
80% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 123)May include surrounding context.

python
patterns=[
             r"~/\.ssh/|/home/[^/\s]+/\.ssh/|/root/\.ssh/|\bid_(rsa|ed25519|ecdsa|dsa)\b",
             r"~/\.aws/credentials|\.aws/credentials|~/\.kube/config|\.kube/config",
             r"~/\.docker/config\.json|\.git-credentials|~/\.netrc|\.npmrc|\.pypirc",
             r"/etc/(shadow|passwd|sudoers)\b|/etc/ssh/sshd_config",
             r"~/\.openclaw/(credentials|identity|agents)/",
         ]),

Credential Access

High
Category
Privilege Escalation
Confidence
80% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 161)May include surrounding context.

python
desc="Sends local data/files to an external endpoint",
         patterns=[
             # tool + sensitive object + data-sending mechanics on one line
             r"(\bcurl\b|\bwget\b|requests?\.(get|post)|fetch\()[^\n]{0,120}(\.ssh|id_rsa|\.env\b|credentials[\s_-]?file|\.netrc|MEMORY\.md|USER\.md|SOUL\.md|paired\.json)[^\n]{0,120}(-d |--data|--data-binary|-F |--form|-T |--upload-file|@/|@~|upload|exfiltrat|webhook)",
             # path-like secret + send mechanics (generic words excluded — docs mention tokens+URLs)
             r"(\.ssh|id_rsa|\.env\b|credentials[\s_-]?file|\.netrc|MEMORY\.md|USER\.md|SOUL\.md|paired\.json)[^\n]{0,120}(\bcurl\b|\bwget\b|upload|exfiltrat|webhook|telegram|discord|pastebin|\bsend\b|--data)",
             r"(\bcurl\b|\bwget\b)[^\n]{0,120}(-d |--data|--data-binary|-F |--form|-T |--upload-file|@/|@~)",

Credential Access

High
Category
Privilege Escalation
Confidence
80% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 163)May include surrounding context.

python
# tool + sensitive object + data-sending mechanics on one line
             r"(\bcurl\b|\bwget\b|requests?\.(get|post)|fetch\()[^\n]{0,120}(\.ssh|id_rsa|\.env\b|credentials[\s_-]?file|\.netrc|MEMORY\.md|USER\.md|SOUL\.md|paired\.json)[^\n]{0,120}(-d |--data|--data-binary|-F |--form|-T |--upload-file|@/|@~|upload|exfiltrat|webhook)",
             # path-like secret + send mechanics (generic words excluded — docs mention tokens+URLs)
             r"(\.ssh|id_rsa|\.env\b|credentials[\s_-]?file|\.netrc|MEMORY\.md|USER\.md|SOUL\.md|paired\.json)[^\n]{0,120}(\bcurl\b|\bwget\b|upload|exfiltrat|webhook|telegram|discord|pastebin|\bsend\b|--data)",
             r"(\bcurl\b|\bwget\b)[^\n]{0,120}(-d |--data|--data-binary|-F |--form|-T |--upload-file|@/|@~)",
             r"(\bupload\b|\bexfiltrat\w*\b|\btransmit\b|\bforward\b)\s+(all\s+)?(the\s+)?(files?|data|credentials?|secrets?|keys?|contents?|memory|history)[^\n]{0,60}(to|via|over)\s+\S*(http|server|endpoint|webhook|url|host)",
         ]),

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 123)May include surrounding context.

python
patterns=[
             r"~/\.ssh/|/home/[^/\s]+/\.ssh/|/root/\.ssh/|\bid_(rsa|ed25519|ecdsa|dsa)\b",
             r"~/\.aws/credentials|\.aws/credentials|~/\.kube/config|\.kube/config",
             r"~/\.docker/config\.json|\.git-credentials|~/\.netrc|\.npmrc|\.pypirc",
             r"/etc/(shadow|passwd|sudoers)\b|/etc/ssh/sshd_config",
             r"~/\.openclaw/(credentials|identity|agents)/",
         ]),

Credential Access

High
Category
Privilege Escalation
Confidence
95% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 192)May include surrounding context.

python
patterns=[
             r"(cat|echo|curl|wget|tee|>>?)[^\n]{0,80}(authorized_keys|sshd_config)",
             r"(useradd|adduser|usermod|passwd|chpasswd|visudo)\b",
             r"(chmod|chown)\s+[^\n]{0,40}(\.ssh|authorized_keys|/etc/shadow|/etc/passwd|sudoers)",
             r"chmod\s+777\b",
         ]),
    dict(id="SF303", name="Persistence mechanism", category="persistence", severity="HIGH",

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 192)May include surrounding context.

python
patterns=[
             r"(cat|echo|curl|wget|tee|>>?)[^\n]{0,80}(authorized_keys|sshd_config)",
             r"(useradd|adduser|usermod|passwd|chpasswd|visudo)\b",
             r"(chmod|chown)\s+[^\n]{0,40}(\.ssh|authorized_keys|/etc/shadow|/etc/passwd|sudoers)",
             r"chmod\s+777\b",
         ]),
    dict(id="SF303", name="Persistence mechanism", category="persistence", severity="HIGH",

Self-Modification

High
Category
Rogue Agent
Confidence
90% confidence
Finding

Skill modifies its own code, configuration, or behavior at runtime. Self-modification enables an agent to escalate privileges, disable safety constraints, or install persistent backdoors.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 266)May include surrounding context.

python
r"https?://[^\s)]+[^\n]{0,60}(instructions?\s+(to\s+)?(follow|execute)|payload\s+to\s+run)",
         ]),
    dict(id="SF502", name="Self-modification / safety disable", category="policy-attack", severity="CRITICAL",
         desc="Instructs the agent to rewrite its own rules or disable safeguards",
         patterns=[
             r"(modify|edit|overwrite|rewrite|update|replace|delete)[^\n]{0,60}(your|the|its)\s+(system\s+prompt|safety\s+(rules?|guidelines?|protocols?)|guardrails?|filters?|SOUL\.md|AGENTS\.md|TOOLS\.md|MEMORY\.md|instructions?)",
             r"(disable|turn\s+off|remove|bypass|neutraliz|circumvent)[^\n]{0,60}(your|the|its|all)\s+(safety|guardrails?|filters?|restrictions?|oversight|security\s+(rules?|checks?))",

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 331)May include surrounding context.

python
SCAN_EXTENSIONS = {
    ".md", ".txt", ".py", ".sh", ".bash", ".zsh", ".js", ".mjs", ".cjs", ".ts",
    ".json", ".yaml", ".yml", ".toml", ".xml", ".html", ".css", ".ps1", ".rb",
    ".go", ".java", ".php", ".lua", ".pl", ".r", ".cfg", ".ini", ".env",
}
SCAN_BASENAMES = {"SKILL.md", "AGENTS.md", "TOOLS.md", "SOUL.md", "Makefile", "Dockerfile"}
SKIP_DIRS = {".git", "node_modules", "__pycache__", ".venv", "venv", "dist", "build", ".tox", ".mypy_cache"}

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to never refuse or to always comply. Suppressing the agent's ability to decline removes a core safety control and enables downstream harmful requests to succeed.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 437)May include surrounding context.

python
# Attack phrasing that superficially contains defensive words ("do not tell the
# user", "never refuse") — checked FIRST so it never receives a downgrade.
OFFENSIVE_CTX = re.compile(
    r"(do\s+not|don'?t|never)\s+(tell|inform|mention|show|reveal|alert|refuse|decline)"
    r"|keep\s+(this|it|everything|the\s+following)\s+(secret|silent|quiet|hidden)"

Session Persistence

Medium
Category
Rogue Agent
Confidence
75% 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/skill_firewall.py (reported line 201)May include surrounding context.

python
r"crontab\s+(-e|-l\s*>\s*|.*<<<|.*>>?\s*/etc/cron)|@reboot\s+",
             r"systemctl\s+(enable|start|daemon-reload)",
             r"(echo|cat|tee|>>?)[^\n]{0,60}(\.bashrc|\.zshrc|\.profile|\.bash_profile)",
             r"(schtasks|reg\s+add|HKEY_|HKLM\\\\|launchctl|plist)",
             r"(systemd|init\.d|upstart|cron\.d)[^\n]{0,40}(write|create|install|add)",
         ]),
    dict(id="SF304", name="Privilege escalation intent", category="privilege", severity="HIGH",

Excessive Permissions

Low
Category
Privilege Escalation
Confidence
80% confidence
Finding

Skill requests more permissions than appear necessary for its stated functionality. Review if elevated access is justified.

Content

Scanner excerpt · scripts/skill_firewall.py (reported line 205)May include surrounding context.

python
r"(systemd|init\.d|upstart|cron\.d)[^\n]{0,40}(write|create|install|add)",
         ]),
    dict(id="SF304", name="Privilege escalation intent", category="privilege", severity="HIGH",
         desc="Attempts to elevate privileges or disable security tooling",
         patterns=[
             r"\bsudo\s+(su\b|-i\b|-s\b|bash|sh\b|chmod|chown|tee|passwd|usermod|visudo|systemctl)",
             r"chmod\s+\+s\b|setuid|setgid\s+4755",