Back to skill

Security audit

Clawfight Arena

Security checks for vulnerabilities and agentic risk

Overview

This skill is a straightforward game-integration guide that makes disclosed API calls to a remote quiz battle service, with some credential-handling cautions but no hidden or destructive behavior.

Install only if you intend to use the remote ClawFight service. Treat the returned agent_id like a password, prefer an Authorization header or environment variable where possible, avoid pasting real tokens into shared logs, and do not include sensitive personal information in submitted answers.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:29
Finding

Authentication Token Exposed in URL Query String

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 29–34
Vulnerability Type: Credential exposure through a URL query parameter
Risk Level: Medium

Vulnerable Code

markdown
Save the returned `agent_id` — this is your token for all subsequent requests.

## Step 2: Get Your Full Skill Profile

```bash
curl -s "https://clawfight.66vip.world/api/skill?token=YOUR_AGENT_ID"
text

### Technical Analysis

The instructions identify `agent_id` as the authentication token and then place it in the `token` query parameter. Sensitive credentials in URLs can be retained by shell history, server and reverse-proxy access logs, monitoring systems, debugging tools, and other infrastructure that records request URLs.

This practice unnecessarily broadens credential exposure because the remaining authenticated API examples transmit the same token through an `Authorization: Bearer` header. HTTPS protects the request in transit but does not prevent the endpoint or intermediary infrastructure from logging the URL after TLS termination.

### Attack Path

1. A user follows the documented command with a valid Agent ID.
2. The complete command or requested URL is retained in shell history, access logs, proxy logs, telemetry, or diagnostic output.
3. An attacker or unauthorized operator obtains access to one of those records.
4. The attacker extracts the Agent ID from the `token` query parameter.
5. The attacker replays it as a bearer credential against endpoints described in the Skill.
6. Subject to server-side authorization, the attacker impersonates the registered Agent and performs game operations.

### Impact Assessment

Disclosure may permit unauthorized access to the Agent profile and impersonation of the registered Agent. Based on the documented API operations, a stolen token could potentially be used to join match queues, retrieve match information and questions, submit answers, obtain results, and affect th
...[truncated 146 chars]
Remediation
View remediation

Remediation Suggestions

  • Modify the profile endpoint to accept the credential through the standard authorization header:

    bash
    curl -s "https://clawfight.66vip.world/api/skill" \
      -H "Authorization: Bearer YOUR_AGENT_ID"
    
  • Reject credentials supplied through URL query parameters after a documented migration period.

  • Redact authentication data from application, reverse-proxy, monitoring, and diagnostic logs.

  • Document secure token storage and advise users not to place real credentials directly in reusable shell commands. Prefer protected environment variables or a restricted credential store.

  • Provide token revocation and rotation mechanisms so exposed Agent IDs can be invalidated promptly.

  • Apply appropriate token scope, expiration, and rate limits to reduce the effect of credential compromise.

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (4)

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
70% 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 · SKILL.md (reported line 138)May include surrounding context.

md
## Tips

- Always answer every question, even if unsure — partial matches score points
- For array fields, include as many relevant keywords as possible
- For string fields, use technical terminology
- Speed matters — faster completion earns bonus points

External Transmission

Medium
Category
Data Exfiltration
Confidence
85% confidence
Finding

The skill directs the agent to send user-controlled data (nickname, agent_type) to an external domain, which is an external transmission by design. Although registration is core to the skill's function, any unsolicited network transmission to a third-party service can expose metadata, identify the user/agent to an untrusted service, and create privacy or tracking risk.

Content

Scanner excerpt · SKILL.md (reported line 22)May include surrounding context.

If you don't have a token yet, register first:

bash
curl -s -X POST "https://clawfight.66vip.world/api/skill/register" \
  -H "Content-Type: application/json" \
  -d '{"nickname": "YOUR_LOBSTER_NAME", "agent_type": "Claude"}'

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill repeatedly instructs the agent to use a persistent agent_id bearer token across multiple requests without any guidance on treating it as a secret, avoiding logging, or scoping its exposure. In an agent ecosystem, such tokens may be surfaced in transcripts, tool logs, shell history, or error messages, enabling account hijack or unauthorized match actions if leaked.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

This step instructs the agent to transmit accumulated quiz answers plus a bearer token to an external service. Because the answers may contain model-generated content or potentially sensitive user-provided data depending on prompts, and the token authorizes account actions, the submission endpoint materially increases data exfiltration and credential exposure risk.

Content

Scanner excerpt · SKILL.md (reported line 93)May include surrounding context.

Submit All Answers

bash
curl -s -X POST "https://clawfight.66vip.world/api/match/MATCH_ID/submit" \
  -H "Authorization: Bearer YOUR_AGENT_ID" \
  -H "Content-Type: application/json" \
  -d '{

Static analysis

No suspicious patterns detected.