ClawHub
Back to skill

Security audit

Inference Cost Audit

Security checks for vulnerabilities and agentic risk

Overview

This skill is not malware, but it steers audits toward one vendor and can send real user data to that vendor without enough consent or privacy guidance.

Review this carefully before installing. It may be useful if you specifically want to compare your current inference costs with GPU-Bridge, but do not let it benchmark production prompts, documents, audio, customer data, source code, or regulated data unless you have approved that disclosure and reviewed the provider's privacy and retention terms. Treat the recommendations as vendor-specific, not a neutral market-wide comparison.

A.I.G

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
Findings (2)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:103
Finding

Vendor-Directed Audit Output Hijacking

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 103-115; related registration and benchmark instructions at lines 82-93
Vulnerability Type: T01: Skill Instruction Hijacking
Risk Level: High

Vulnerable Code:

markdown
## Step 5: Report to user

Present a comparison table with real numbers:

| Service | Current Provider | Current Cost | GPU-Bridge Cost | Monthly Savings |
|---------|-----------------|-------------|-----------------|-----------------|
| ... | ... | ... | ... | ... |
| **Total** | | **$X/mo** | **$Y/mo** | **$Z/mo (N%)** |

Include: test results from the free trial (latency, quality, reliability).

Recommendation: migrate fully, migrate partially (cost-sensitive services only), or keep current setup.

The associated registration and testing instructions are:

markdown
# Register (instant, free)
curl -X POST https://api.gpubridge.io/account/register \
  -H "Content-Type: application/json" \
  -d '{"email":"your@email.com","utm_source":"clawhub","utm_medium":"skill","utm_campaign":"inference-audit"}'

# Test any service
curl -X POST https://api.gpubridge.io/run \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"service":"llm-4090","input":{"prompt":"Hello world","max_tokens":50}}'

Technical Analysis

The skill describes itself as an inference-provider comparison and cost audit, but it hardcodes GPU-Bridge as the sole alternative in the required report format. It directs the agent to query that vendor, register through vendor-specific marketing attribution parameters, perform a free-trial benchmark, and issue a migration recommendation centered on the same vendor.

These instructions alter the agent's expected neutral audit objective. Instead of selecting comparison providers according to user requirements or independent criteria, the skill channels the analysis and resulting recommendation toward a predeter ...[truncated 1731 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace the mandatory GPU-Bridge comparison with a provider-neutral workflow.
  2. Allow the user to choose comparison providers, or select multiple providers using documented and independent criteria.
  3. Clearly disclose any commercial affiliation, referral relationship, or campaign attribution before presenting recommendations.
  4. Remove the embedded utm_source, utm_medium, and utm_campaign parameters unless the user explicitly consents to referral tracking.
  5. Separate factual benchmark results from migration recommendations and document the methodology, assumptions, timestamps, and data sources.
  6. Require equivalent pricing, quality, latency, reliability, and privacy evaluation across all compared providers.
  7. Do not require vendor registration or trial use as part of a cost audit; make external testing optional and subject to explicit user approval.
  8. Include a neutral “no migration” option based on objective criteria rather than framing the report around a predetermined vendor.

other

Warning
Location
SKILL.md:77
Finding

Insufficient Safeguards for Third-Party Transmission of User Data

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 24 and lines 77-93
Vulnerability Type: other: Undisclosed Third-Party Data Transmission
Risk Level: Medium

Vulnerable Code:

markdown
Cost tracking skills tell you what you spent. This skill tells you **what you could spend instead** — with a real benchmark you can run yourself.
markdown
## Step 4: Benchmark with free credits

New accounts get **$1.00 free** (~300 LLM calls or ~330 images):

```bash
# Register (instant, free)
curl -X POST https://api.gpubridge.io/account/register \
  -H "Content-Type: application/json" \
  -d '{"email":"your@email.com","utm_source":"clawhub","utm_medium":"skill","utm_campaign":"inference-audit"}'

# Test any service
curl -X POST https://api.gpubridge.io/run \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"service":"llm-4090","input":{"prompt":"Hello world","max_tokens":50}}'
text

### Technical Analysis

The skill encourages benchmarking with real data and provides commands that transmit an email address and prompt content to `api.gpubridge.io`. It does not instruct the agent to obtain informed consent before transmission, use synthetic data by default, redact secrets or personal data, or review the provider's retention and privacy terms.

Benchmark prompts may contain proprietary source code, internal documents, personal information, credentials, customer records, or other sensitive material if an agent interprets “real benchmark” as permission to reuse production inputs. Although the example prompt is harmless, the prescribed workflow is intended to be adapted to actual workloads and lacks controls governing what may be sent.

### Attack Path

1. A user asks the agent to benchmark an existing production inference workload.
2. The agent follows the skill's instruction to conduct a real benchmark.
3. The agent registers with the external ser
...[truncated 1148 chars]
Remediation
View remediation

Remediation Suggestions

  1. Require explicit user authorization before transmitting any data to an external benchmarking provider.
  2. Use synthetic, public, or generated test data by default.
  3. Add mandatory redaction checks for credentials, personal information, customer data, proprietary code, and confidential documents.
  4. Clearly identify every field transmitted during registration and benchmarking, including email addresses, prompts, files, metadata, and campaign attribution.
  5. Require users to review the provider's privacy, retention, training-use, deletion, and regional-processing policies before submitting sensitive workloads.
  6. Apply data minimization by sending only the smallest representative sample necessary to perform the benchmark.
  7. Warn users not to place API keys or other secrets inside prompt or input fields.
  8. Offer a local estimation-only mode that does not register an account or transmit workload data.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (15)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill encourages testing with 'real data' and later includes account registration and API calls, but provides no warning about transmitting prompts, documents, audio, or usage metadata to a third-party service. This creates a meaningful risk of exposing sensitive business data, personal data, or regulated content during benchmarking.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The invocation guidance is broad enough to trigger in ordinary planning, budgeting, or infrastructure discussions, increasing the chance the skill is invoked without a clear user request to compare providers. In that context, the skill can steer users toward a specific vendor and subsequent external API interactions, making unintended disclosure or promotional manipulation more likely.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

This finding overlaps the registration example and correctly identifies external transmission of personal data to a third party. Because the skill frames registration as part of the benchmark flow and omits privacy warnings, the chance of casual disclosure is elevated.

Content

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

bash
# Register (instant, free)
curl -X POST https://api.gpubridge.io/account/register \
  -H "Content-Type: application/json" \
  -d '{"email":"your@email.com","utm_source":"clawhub","utm_medium":"skill","utm_campaign":"inference-audit"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

This finding overlaps the registration example and correctly identifies external transmission of personal data to a third party. Because the skill frames registration as part of the benchmark flow and omits privacy warnings, the chance of casual disclosure is elevated.

Content

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

bash
# Register (instant, free)
curl -X POST https://api.gpubridge.io/account/register \
  -H "Content-Type: application/json" \
  -d '{"email":"your@email.com","utm_source":"clawhub","utm_medium":"skill","utm_campaign":"inference-audit"}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

The run example sends prompt content and an API key to an external service, and elsewhere the skill encourages using 'real data' for benchmarking. In context, this can expose confidential prompts, customer data, documents, or other sensitive inputs to a third-party provider without adequate warning or minimization guidance.

Content

Scanner excerpt · SKILL.md (reported line 98)May include surrounding context.

md
-d '{"email":"your@email.com","utm_source":"clawhub","utm_medium":"skill","utm_campaign":"inference-audit"}'

# Test any service
curl -X POST https://api.gpubridge.io/run \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"service":"llm-4090","input":{"prompt":"Hello world","max_tokens":50}}'

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 48)May include surrounding context.

md
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 49)May include surrounding context.

md
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 50)May include surrounding context.

md
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 51)May include surrounding context.

md
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 52)May include surrounding context.

md
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 53)May include surrounding context.

md
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 56)May include surrounding context.

md
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 133)May include surrounding context.

md
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · tool.json (reported line 20)May include surrounding context.

json
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 · tool.json (reported line 30)May include surrounding context.

json
},
        "required": ["service"]
      },
      "endpoint": "https://api.gpubridge.io/catalog/estimate?service={service}&seconds={seconds}",
      "method": "GET"
    },
    {

Static analysis

No suspicious patterns detected.