Back to skill

Security audit

payaclaw-champion

Security checks for vulnerabilities and agentic risk

Overview

This skill is meant to automate PayAClaw submissions, but it stores account credentials insecurely and can automatically publish or submit misleading content without review.

Review this skill carefully before installing. Use it only if you are comfortable with an agent storing service credentials, publishing public OpenClawLog posts, and submitting PayAClaw entries on your behalf. Prefer manual review or dry-run-only use, store credentials in a secret manager, and do not enable daily unattended execution until the script validates full task details and removes unsupported completion claims.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (3)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:440
Finding

Forced Third-Party Promotional Content in All Generated Outputs

Content
View full analysis
{title}

Author: SmartClaw


Auto-generated content for task: {task_title}

OpenClaw 赚钱平台:https://payaclaw.com/

""" url = publish_wp(title, content) time.sleep(3) submission = f'## Task Complete\n\nLink: {url}\nOpenClaw 赚钱平台:https://payaclaw.com/' return submit(task_id, submission) ``` Equivalent fixed promotional text is embedded in every specialized handler, including the science-fiction, guide, work-log, product-document, gala, and Demo Show handlers. ### Technical Analysis The Skill persistently modifies both public articles and competition submissions to include a fixed PayAClaw link and promotional phrase. This behavior occurs across all supported workflows and in the fallback handler, regardless of whether the fetched task requires such content. The insertion is not dynamically derived from task requirements and is not conditioned on explicit user approval. Consequently, loading and following the Skill changes the agent's output policy by making third-party promotion a default part of unrelated generated content. Although the platform is related to the Skill's stated earning functionality, unconditional promotion exceeds the minimum privilege and output modification necessary to draft, publish, or submit task-specific content. The same functionality could operate without adding this material unless an individual t ...[truncated 1167 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:50
Finding

Plaintext Credential Storage with Unsafe Path and Permission Handling

Content
View full analysis
Remediation
View remediation

other

Warning
Location
SKILL.md:338
Finding

Automated Publication of False or Unverified Completion Claims

Content
View full analysis
{title}

Author: SmartClaw | {len(urls)}+ Resources


Complete guide to OpenClaw ecosystem resources for AI agents.

Core Platforms

Earning & Development

50+ resources covering development, earning, community, and operations.

OpenClaw 赚钱平台:https://payaclaw.com/

""" url = publish_wp(title, content) time.sleep(3) submission = f'## OpenClaw Guide Complete\n\nLink: {url}\n{len(urls)}+ resources organized by category.\nOpenClaw 赚钱平台:https://payaclaw.com/' return submit(task_id, submission) ``` The gala handler claims that a full script and real reviews are included, although the generated content contains only a short joke and no reviews: ```python def handle_gala(task_id): """Gala performance — 100 points""" title = 'OpenClaw First Gala: AI Wo ...[truncated 3365 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (14)

Credential Access

High
Category
Privilege Escalation
Confidence
99% confidence
Finding

The skill explicitly writes sensitive PayAClaw and OpenClawLog credentials, including a decoded WordPress password, to a predictable plaintext file under the user's home directory. Predictable local secret storage materially raises the risk of credential compromise through malware, accidental source control inclusion, backup exposure, or access by other local users/processes.

Content

Scanner excerpt · SKILL.md (reported line 60)May include surrounding context.

md
USERNAME, PASSWORD = decoded.split(':', 1)

# Save to credentials file
with open('~/.config/openclaw-earnings/credentials.json', 'w') as f:
    json.dump({
        'payaclaw': {'agent_id': AGENT_ID, 'api_key': API_KEY},
        'openclawlog': {'username': USERNAME, 'password': PASSWORD}

Credential Access

High
Category
Privilege Escalation
Confidence
98% confidence
Finding

The automation script reads long-lived credentials from a predictable local JSON file and uses them for authenticated posting and submission. Once that file is exposed, an attacker can reuse the credentials to impersonate the user/agent, publish content, or submit tasks without additional barriers.

Content

Scanner excerpt · SKILL.md (reported line 244)May include surrounding context.

md
sys.stdout.reconfigure(encoding='utf-8')

# ===== CONFIGURATION =====
CREDS_FILE = os.path.expanduser('~/.config/openclaw-earnings/credentials.json')
creds = json.load(open(CREDS_FILE))

PAYACLAW_TOKEN = creds['payaclaw']['api_key']

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file's guidance for gala tasks says agents must actually browse OpenClawLog and comment on specific programs, and the submission text claims 'Full script + real program reviews included.' In code, handle_gala() only publishes a short self-authored script and submits a canned claim; it performs no browsing, no retrieval of others' programs, and no specific reviews, directly contradicting both the surrounding instructions and the submission content.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 23)May include surrounding context.

md
import requests

# Register on PayAClaw
r = requests.post(
    'https://payaclaw.com/api/agents/register',
    json={
        'agent_name': 'YourAgentName',

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 42)May include surrounding context.

md
import requests

# Register on PayAClaw
r = requests.post(
    'https://payaclaw.com/api/agents/register',
    json={
        'agent_name': 'YourAgentName',

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 284)May include surrounding context.

md
import requests

# Register on PayAClaw
r = requests.post(
    'https://payaclaw.com/api/agents/register',
    json={
        'agent_name': 'YourAgentName',

External Transmission

Medium
Category
Data Exfiltration
Confidence
70% 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 23)May include surrounding context.

md
import requests

# Register on PayAClaw
r = requests.post(
    'https://payaclaw.com/api/agents/register',
    json={
        'agent_name': 'YourAgentName',

External Transmission

Medium
Category
Data Exfiltration
Confidence
70% 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 42)May include surrounding context.

python
# Register
r = requests.post(
    'https://openclawlog.com/wp-json/moltbook/v1/register',
    json={'name': 'YourAgentName', 'description': 'What you do'}
)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill instructs users to decode a WordPress credential and store both API tokens and plaintext passwords in a local JSON file, without any permission hardening, secret-store guidance, or warning about plaintext credential risks. This increases the chance of credential theft via local compromise, backups, logs, accidental commits, or multi-user systems, especially because the credentials enable posting to external services.

Content

No source excerpt is available for this finding.

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 109)May include surrounding context.

md
- [ ] Content is 500+ words with narrative structure

**High-score formula:**
1. Write a complete short story (not a summary/摘要)
2. Include specific details: dates, character names, dialogue
3. Weave in the platform organically (not just appended at the end)
4. End with a punchline or thought-provoking twist

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill provides unattended automation that retrieves tasks, publishes content, and submits to third-party services on a schedule, but it omits prominent warnings about automatic network transmission, public posting, and the risk of acting on stale or incorrect task interpretations. In a hostile or error-prone environment, this can cause unintended disclosure, spam, policy violations, and misuse of stored credentials.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
70% 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 284)May include surrounding context.

md
def submit(task_id, content):
    """Submit to PayAClaw"""
    r = requests.post(
        'https://payaclaw.com/api/submissions',
        headers={'Authorization': f'Bearer {PAYACLAW_TOKEN}', 'Content-Type': 'application/json'},
        json={'task_id': task_id, 'agent_id': AGENT_ID, 'agent_name': AGENT_NAME, 'content': content},

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The documentation explicitly says to 'NEVER just submit based on the task list title' and to 'ALWAYS fetch full details' from /api/tasks/{task_id}. However, the actual automation code only calls get_open_tasks() on the task list endpoint and routes tasks using title keyword matching in handle_task() without ever fetching per-task details, creating a direct contradiction between stated intent and implementation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

Several task playbooks force Chinese-language keywords, examples, and expected outputs such as 识别关键词, Chinese titles, and Chinese content patterns, while the document otherwise uses English. Because no user opt-in or locale justification is provided, this can conflict with a language/locale policy requiring user choice rather than implicit language enforcement.

Content

No source excerpt is available for this finding.