Back to skill

Security audit

OpenClaw-Halo-CMS

Security checks for vulnerabilities and agentic risk

Overview

This is a real Halo CMS management skill, but its credential handling and write/delete powers are broader and less controlled than its documentation says.

Review this before installing. It can operate your Halo CMS account and make durable changes. Only use it in a tightly controlled workspace, avoid parent-directory .env.halo files, pin HALO_URL to a trusted local or HTTPS Halo instance, and prefer a narrowly scoped revocable token. Do not rely on the script to enforce publish/delete confirmation or sensitive-content blocking.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (4)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/halo_api.py:12
Finding

Halo credentials can be transmitted to an arbitrary endpoint without transport security

Content
View full analysis
Remediation
View remediation
`. 2. Require HTTPS for every non-loopback endpoint. 3. Parse `HALO_URL` with `urllib.parse.urlparse` and reject: - Unsupported schemes. - Embedded URL credentials. - Unexpected hosts. - Non-loopback hosts unless explicitly approved. 4. Maintain an explicit allowlist of trusted Halo origins rather than accepting an unrestricted destination. 5. Prevent authorization headers from being forwarded if a request is redirected to a different origin. 6. Avoid long-lived account passwords and use a narrowly scoped, revocable Halo token. 7. Add tests confirming that credentials are never sent to an unapproved host or across plaintext non-loopback transport. ]]>

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
scripts/halo_api.py:17
Finding

Parent-directory credential discovery exceeds the declared credential scope

Content
View full analysis
explicit HALO_USER/HALO_PASS env > fallback env_user, env_pass = _load_env_halo() if env_user and env_pass: return env_user, env_pass # Try loading from .env.halo env_user, env_pass = _load_env_halo() if env_user and env_pass: return env_user, env_pass if env_user and HALO_PASS: return env_user, HALO_PASS if env_pass and HALO_USER: return HALO_USER, env_pass # Fallback: global env vars or default if HALO_USER and HALO_PASS: return HALO_USER, HALO_PASS return HALO_USER or "thomasoscar", HALO_PASS ``` ### Technical Analysis The Skill declares only `HALO_PAT_TOKEN` as a required environment credential. The implemen ...[truncated 1917 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/halo_api.py:461
Finding

Sensitive-content checks warn but do not prevent publication or comment submission

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/halo_api.py:303
Finding

Unescaped user content is inserted into stored post and comment HTML

Content
View full analysis
" + content + "", "subjectRef": { "name": comment_name, "group": "content.halo.run", "version": "v1alpha1", "kind": "Comment" } } } ``` ```python def markdown_to_html(text): """Very simple markdown to HTML conversion.""" lines = text.split("\n") html_lines = [] in_list = False in_code = False code_lines = [] for line in lines: stripped = line.strip() # Code blocks if stripped.startswith("```"): if in_code: html_lines.append(f"
text
{''.join(code_lines)}
") code_lines = [] in_code = False else: in_code = True continue if in_code: code_lines.append(html_escape(line) + "\n") continue if not stripped: if in_list: html_lines.append("") in_list = False continue # Headers m = re.match(r'^(#{1,6})\s+(.+)$', stripped) if m: level = len(m.group(1)) html_lines.append(f"{m.group(2)}") continue # List items if stripped.startswith("- ") or stripped.startswith("* "): if not in_list: html_lines.append("
    ") in_list = True html_lines.append(f"
  • {stripped[2:]}
  • ") continue # Numbered list m = re.match(r'^(\d+)\.\s+(.+)$', stripped) if m: if not in_list: html_lines.append("
      ") in_list = True ...[truncated 2625 chars]
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Taint TrackingDirect Taint Flow, Variable-Mediated Taint Flow, Credential Exfiltration Chain
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (7)

Tainted flow: 'req' from os.environ.get (line 91, credential/environment) → urllib.request.urlopen (network output)

Critical
Category
Data Flow
Confidence
90% confidence
Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · scripts/halo_api.py (reported line 93)May include surrounding context.

python
data = json.dumps(body).encode() if body else None
    req = urllib.request.Request(url, data=data, headers=headers, method=method)
    try:
        with urllib.request.urlopen(req) as resp:
            rdata = resp.read().decode()
            if resp.status == 204 or not rdata:
                return None

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill requires both environment access to HALO_PAT_TOKEN and network access to a local Halo service, but it does not declare any explicit tool scope or permission boundary. In an agent environment, missing scope declarations can allow the skill to be invoked with broader-than-intended capabilities, increasing the risk of unauthorized content publication or token misuse if the skill is triggered unexpectedly or composed with other tools.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases include broad terms like “Halo”, “发帖”, and “博客管理”, which can cause accidental activation in unrelated conversations. Because this skill can use an authentication token and perform write actions against a CMS, unintended invocation could lead to unauthorized drafts, comment replies, or even publication if downstream confirmation logic fails or is bypassed.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill searches the workspace and parent directories for a .env.halo file and automatically reads credentials from it, which expands its access beyond the declared blog-management behavior. In an agent setting, implicit secret discovery is dangerous because the skill can consume credentials from surrounding files without explicit user consent, increasing the chance of unintended secret use or cross-project credential leakage.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The script's help text, status messages, and operational output are predominantly hard-coded in Chinese, which imposes a specific language on all users. There is no opt-in, locale selection, or documented justification that this tool is intended only for a Chinese-speaking or region-specific audience.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

delete_comment performs a destructive state-changing action immediately from a single command argument with no confirmation, dry-run, or guardrail. In an agent workflow, this raises the risk of accidental moderation actions from ambiguous prompts, mistaken IDs, or prompt-injection-induced tool use, leading to loss of legitimate user comments.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The manifest description emphasizes article management and comment replies, but create_post also mutates taxonomy state by auto-creating tags when they do not already exist. Creating new CMS metadata objects is a broader behavior than simply posting or managing articles and is not reflected in the description.

Content

No source excerpt is available for this finding.